Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Mining Educational Data to Analyze Students’

Performance: A Case Study of Mawuli School, Ho
Cephas Lokpo
Mathematics/ICT Department
Akatsi College of Education
Akatsi, Ghana

Abstract:- Data mining is the extraction of prospective I. INTRODUCTION

valuable information from large chunk of data through
the employment of many different data mining Predicting students’ scholastic accomplishment is
techniques. The usefulness of data mining coupled with critical for monitoring progress in establishments of
the huge data generated in scholastic settings has made it learning. The scholastic output of students is centered on
an interesting field of research known as Educational some varying issues namely psychosomatic, social, personal
Data Mining (EDM). The intent of EDM is to derive and other environmental agents. Data Mining is a very
understanding from hidden patterns in data collected useful tool that can be employed in helping to achieve
from institutions of learning to aid in identifying issues objectives of scholastic accomplishment. Tools used in
that influence students’ scholastic accomplishment, to extracting knowledge have the capability of maneuvering
solution of which will lead to improvement in volumes of sets of data to determine unseen samples besides
accomplishment. Because scholastic achievement is being helpful links in decision making (Bharadwaj and Pal,
dependent on several issues, it is essential to develop 2011). Prediction models have a definite goal that enables us
predictive models on students’ academic performance. conjecture unspecified estimates of the observed variables
This study’s objective, therefore, is to acquire an insight given recognized values of other variables. In predicting the
into performance through knowledge discovery by the accomplishment of the learner effectively, it is necessary to
use of simple linear regression in order to build a use plans such as the earlier mentioned variable elements.
predictive model capable of predicting students’ grades The benefit of predicting the accomplishment of students
to give a general overview of students’ performance in with precision helps in identifying students who perform
the WASSCE, and help improve students’ performance. low academically in the beginning, and the subject areas that
In accomplishing the set objective for which the study students need further assistance. It is essential for the
was carried out, that is to predict the possible outcome of teacher to give more assistance to such students to help
students in WASSCE, a widely sampled study was ameliorate their scholastic accomplishment in the future.
applied in undertaking the study (Quantitative
research). The case study chosen was Mawuli School The integration of IT in diverse disciplines of human
Correlation and Regression was employed to determine endeavor has led the abundant storage of data in data
the degree to which variables are connected and to repositories in different formats. Data collection from
construct a model to predict students’ final scores. It was diverse applications necessitates appropriate technique for
revealed that there was a decline in students’ digging up information from a vast data warehouse for
performance in Mathematics, English Language, and informed decisiveness. Mining of data, also known as
Integrated Science, and a good number of students fell Knowledge databases discovery or (KDD), focuses on
below the pass mark, although some few students were discovering valuable information in large sets of data.
performing high, whiles Social Studies saw a good Furthermore, the Core tasks of mining of data include
improvement in students’ performance over utilizing diverse techniques as well as processes which is to
performance in the first year by recording a 98.9% pass. uncover and dig up models of amassed data.
It was also discovered that there is a close association
between students’ first year and second year The process of extracting informative knowledge
performance, and that the effect of students’ first year which includes samples, relations, variations, irregularities
performance on their second year performance was not and considerable organizations obtained from vast
due to likely random chance but that the effect had distribution of information deposited in data warehouses,
underlying repeatable cause, and hence models were databases, or any data storehouses is referred to as Data
constructed to predict students’ final WASSCE mining (DM).(Han and Kamber, 2001). It is regarded as one
scores/grades. of the knowledge discovery steps which interrelate with
knowledge or user base. Patterns of interest are offered to
Keywords:- Educational data mining, performance, users, or maybe kept as fresh information in the database.
association, prediction. Even though it is just one stage of the discovery of
knowledge processes it is very important since unseen
models are discovered for assessment (Han and Kamber,

IJISRT20MAY635 www.ijisrt.com 1920

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
2000). Educational Data Mining (EDM), deriving The grades of students based on performance are not
knowledge from patterns and trends uncovered from in any way predictable in this modern system of education.
learning institutions is a field of research of contemporary Most of the renowned second cycle institutions in Ghana,
time which is concerned for fashioning devices that uncover and most especially in the Volta Region of Ghana, have seen
information from educational databases (Han and Kamber, a serious decline in performance in recent times. The results
2000). of 2013revealed a sharp fall in performance. For instance, in
one of such schools, out of 1089 candidates who took the
These records can be retrieved from archival and examination, 599 students passed in all eight subjects,
active data catalogues in educational institutes. These representing 55% and only 471 qualified for tertiary level,
collected data are either of educational or individual’s representing 43.5%. This decline is so sharp when compared
particulars. Furthermore, e-learning provides large volumes to 2011 and 2012 results that recorded 90.6% and 89.3% of
of data that can be utilized (Romero and Ventura, 2007). candidates that qualified for tertiary level.

Analyzing data gathered from scholastic In spite of this situation, stakeholders always expect
establishments through the use of DM systems to solve high performance even though the stock of candidates
challenges associated with research in education is termed presented for WASSCE may not be the best, and most often
EDM. The purpose of undertaking EDM is to have a better results of poor performance hit them by surprise. Perception
knowledge of students and their place of learning. among some stakeholders is that the stock of students being
Therefore, EDM’s center of attention is gathering, storing admitted through the Computerized School Selection and
and interpretation of facts and figures from students’ studies Placement System (CSSPS) to a large extent is affecting
and evaluation. performance negatively.
Various methods are applied in undertaking EDM
which include Naïve Bayes, Decision Trees, Nearest B. Purpose of the Study
neighbor, Neural Networks, K- Regression, Correlation, etc. The study intended to gain insight into students’
performance through knowledge discovery in order to
Employing methods like classifications, clustering and establish a predictable system capable of predicting
association rules can discover a lot of information that students’ grades in WASSCE in order to give an overview
would be useful in predicting student enrolment for a given of the general performance of students and better inform
course, detecting electronic exam malpractices, errors in management and other stake holders with regard to their
student reports, performance of students and many more. decisions and expectations. This will also help identify how
From the perspective of learning, standardized methods for weak or high students’ grades may be. These projected
measuring the ability and qualities of the human mind which grades will help determine the amount of remedial efforts
lack precision can find a replacement of EDM. At present, a that need to be expended by teachers to help students
good deal of EDM study areas well accommodated in the improve their performance before their final examination to
environment of Intelligent Tutoring System (ITS) because of reduce failure rate.
the greater number of researchers in the EDM domain
focusing attention on mining data originating from ITS. C. Objectives of the Study
1. Analyze the performance of students in the subjects
EDM searches for a better understanding of the namely Integrated Science, Mathematics, English
learning process and the participation of students of it, Language, and Social Studies, using linear regression.
always searching for a quality improvement and cost- 2. Verify the extent to which the grade of students in Basic
effectiveness of the education system. Educational Data Education Certificate Examination (BECE) is connected
Mining has the following objectives (Romero and Ventura, to the grade of students in Senior High School (SHS).
2007): 3. Verify the level of connection between first and second
 Pedagogical: to help in the design of didactic contents year performances of students in all core subjects.
and learners’ scholastic accomplishment promotion; 4. Construct a predictive model for projecting students’ final
 Managerial: optimization of the establishment and grade in all core subjects (Integrated Science,
preservation of structures of training; Mathematics, English Language, and Social Studies)
 Commercial: enrollment of students in institutes of using linear regression analysis.
learning that are owned by individuals nongovernmental
groups in particular. D. Research Questions
1. Do all students perform above average (50%) in
A. Statement of Problem Integrated Science, Mathematics, English Language, and
In Ghana’s educational system, students’ performance Social Studies?
in the senior high school is defined by summing up of the 2. Do students’ grades in junior high school exam (BECE)
marks obtained from examination set by an independent determine their performance in senior high school and to
body and marks accumulated from terminal exams and class what extent?
work. A student is successful when he or she excels above 3. Do students’ terminal performance at SHS determine
the minimum academic requirements both in school and in their final WASSCE grades and to what extent?
WASSCE.

IJISRT20MAY635 www.ijisrt.com 1921

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
E. Significance of the Study A. Mining of Data
Undertaking DM proffers extensive accurate tools for The discipline associated with the discovery of
making decisions, When applied in education, DM tasks possibly valuable data from a vast distribution of data is
indubitably offer broad, yet precise, decision-making tools, called data mining or Knowledge Discovery in Database
detection as well as projection of the consequence of gaining (KDD). Fields of application include marketing, fighting
knowledge and response, and learners’ rating forecasting. terrorism and management of biological information. The
Others include detection of learners who have the possibility urge to undertake scientific investigations in the field of
of not being retained, learner counseling; recognition of EDM has soared up considerably in contemporary time. Of
learners who are excelling, poor-performing students, late attention is drawn to the utilization of techniques used
learners with the same learning characteristics and way of to mine data to examine technical problems in research areas
thinking, disagreeable detecting learner demeanor detection, in education; this is commonly called Educational Data
recognition of learners who are apt to team work, assigning Mining.
the right learning task to the right group of learners,
preparation and drawing timetable for lessons and other The term used to describe a rise in the volume of
learning behaviors, etc. (Sargenti et al., 2006; Rajagopalan information stored in an establishment’s database as well as
and Shyamala, 2006). probability to uncover depths in analyzed is “big data.”
According to IBM, “big data” covers three sundry scopes,
In view of the above quote, this research will in no namely volume, velocity, and variety (IBM, 2012).
small way help in resolving the challenge of low
performance of students because the predictions will reveal Filtering to discover useful information for decision
the nature of the final result and the subject areas where making is a very big challenge to organizations. One
more attention must be given. The highest, lowest, and the efficient instrument to help organizations a great deal to
average marks predicted by the regression model speak a lot unearth important information for decision making at all
about the overall performance of students. The study is levels of management operations is data mining (Kiron et
therefore relevant to management and teaching staff of al., 2012).
Mawuli School for decision making, as it gives the overview Mining of Data involves the use of categories of tools
of students’ performance in their final examination. and techniques to uncover unrecognized samples and
Stakeholders will get the general overview of students’ associations found in the data (Dunham, 2003). DM
projected final results. similarly is by and large a single stage in the process of
discovering information, in which institutions desire to be
II. LITERATURE REVIEW well informed in taking decision using fresh data. It can
therefore be said that in taking major decisions, DM as well
Interests have grown in mining educational data and it as discovery of information plays a crucial and inevitable
is all about formulating techniques to unearth information role.
from educational settings (Garcia et al., 2008).
Employing diverse methods of DM like classifications,
Data is made available in archives (historic data), and clustering and association rules can be used to discover a lot
from information systems on day-to-day transactions of information, as such techniques measurably evaluates
(operative data) found in repositories of educational enormous amount of information to discover unseen designs
establishments. Information of students gathered on and interpretations of information. DM therefore confirms
demographic variables (personal data) or on performance evaluations as well as explains the process (Berson et al.,
variables (academic data). Similarly, electronic systems of 2011).Being highly exploratory, DM varies from
learning provide such information (Romero and Ventura, investigative and analyzing procedures, as it is typically
2007). determined by the problem and confirms.

Higher educational establishments seek to offer its Areas where DM had is employed in various
students with education of the highest quality and advance organizations, in defense as well as in commerce however, it
their decision-making ability. And this can be achieved by has been underutilized in the area of education.(Malik and
unearthing information from their academic data to evaluate Ranjan, 2007).The story is different today; however, as data
the issues that influence learners’ scholastic mining has seen a significant application in the field of
accomplishment. The extracted knowledge is beneficial in education. This buttresses the point Chakrabarti and others
contributing to decision making in the administrative level made that in recent times that data mining is now greatly
as it is productive reference decision makers. Some benefits considered both in education and business (Chakrabarti et
it provides include improving the academic output of al., 2009).
students, reduce poor performance, good comprehension of
student conduct and improve the overall educational process DM in education is an aspect of evaluating and
(Chadha and Kumar, 2011). Educational data mining utilizes applying DM in solving challenges pertaining to learners
numerous methods in discovering data employing and their settings in which they learn (educationally-related
techniques for instance association rules, clustering and problems).Modern techniques in unearthing designs as well
classification. These include decision tree, rule induction, as developments in diverse sets of information in education
neural networks and so on (Romero and Ventura, 2007).

IJISRT20MAY635 www.ijisrt.com 1922

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
is made available to experts in the research field when DM Stages in the KDD method are represented below:
is applied in this way.

From the ongoing discussion on concepts of data

mining, it can be deduced that it is a process of information
extraction from a mass of knowledge, in agreement with
(Han and Kamber, 2001), which stated that: DM is a method
of unearthing fascinating information deposited in huge
storage systems of any kind such as a repository of data,
data store houses. Fascinating data as in one unseen by any
normal technique but of immerse importance.

Considering the import of the definition and

explanation of data mining by the various authors, all of
them seemed to be dwelling on the same idea. The pivotal
issue in all definitions is the discovery of valuable
information from patterns, trends, and relationships
extracted from huge or complex data sets with a primary
purpose of making informed decisions that will improve
efficiency. Figure 2.1: DM: KDD procedure (The steps of extracting
knowledge from data)
Data mining encompasses various fields of computer Source: Baradwaj and Pal (2011)
science. The problem at hand which require a solution as
well as the trend of data of interest, are the factors that The following are areas suitable for data mining:
determine the type of data mining technique to use.  Prediction- estimate verification of unknown attribute on
However, as determined by the situation at hand numerous the basis of designs inherent in the data.
procedures could be approached such as learning aided by  Classification- categorization of facts and figures
machines(Witten and Frank, 1999); man-made intellect, data according to characteristics.
extraction and organization (Mitra and Acharya, 2003); and  Clustering- identification of what is similar and
many more. It can also be said that, various discovering dissimilar in features of a set of data for discovering
strategies yield different trends and models of data these homogeneous/heterogeneous groups that will portray the
may include analyzing in clusters, analyzing by trend, etc. records.
(Han and Kamber, 2001).  Description- interpretation of patterns in the data into
forms that human beings can understand (Fayyad, 1996).
By a careful examination of what techniques, one can
employ in data mining it is obvious from the previous DM functions similarly to the brain of humans. It
paragraph that there are many techniques as suggested by learns new knowledge based on historical information or
(Mitra and Acharya, 2003; Chakrabarti et al., 2009). The experience. That notwithstanding, for data mining
choice of particular technique however, is determined by the technology to arrive at new knowledge (information) in the
kind of problem the researcher is seeking to solve and the database, the user must present to it the patterns. DM
type of pattern and data to be mined as proposed by (Witten utilizes these pieces of data from the past to generate a
and Frank, 1999). In a similar manner, (Han and Kamber) model that will predict what may occur with time by using
suggested that various discovering strategies yield different such information characteristics to identify similar ones in
trends and models. data storage.

The general means of discovering knowledge by the To build predictive or inference models whose aim is
use of data mining techniques involves an interactive order to forecast or project future tendencies or behaviors
as follows: i) Preprocessing step which involves data established on the examination of prearranged data is a
cleaning and amalgamation ii)modeling step which involves function of data mining (Han and Kamber, 2001).
selecting data, transforming it, structuring, trend valuation, Prediction as used contextually means the building of
and information presentation (Han and Kamber, 2006). Data structures that is used to evaluate the kind of unidentified
is analyzed for the purpose of removing noise, missing instance. Alternatively, it means to examine or analyze the
values and inconsistencies before the processing stage. The estimate or value ranges of a character trait that a given
resultant preprocessed is deposited in data storage bases. instance probably has. Two mainly used predicting methods
Several approaches for example, recognizing of patterns are Classification which is exercised in predicting distinct
‘dig out’ and assess configurations in data during the figures and regression which is applied in predicting well-
modeling step. The relevant trends discovered could be ordered data constitute two principal techniques for
delivered to customers through visualizing and presentation forecasting. Classification is the method employed in
methods (Chakrabarti et al., 2009). predicting discrete or nominal estimates and method of
regression utilized for projecting constant estimates (Larose,
2006).

IJISRT20MAY635 www.ijisrt.com 1923

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
B. Educational Data Mining Perhaps in future a broader perspective can be used in
Educational data mining (also referred to as “EDM”) is researches that seek to evaluate the growth of this discipline.
defined as the area of scientific inquiry centered on the
development of methods for making discoveries within the EDM spans spheres of dramatic influence on learners.
unique kinds of data that come from educational settings, Further themes that EDM encompasses have to do with
and using those methods to understand students and the analyzing didactic procedures such as enrollment, old
settings which they learn better. student associations, and choice of program. In addition,
there are unique tools for mining data gathered from
Educational data mining methods often differ from academic institutes and can be functional in extraction of
methods of the broader data mining techniques in literature. data online, categorization, mining to discover rules of
They are methods explicitly used for exploiting the multiple connections, and analysis of data that consist of many
levels of meaningful hierarchy in educational data. To variables (Calders and Pechenizkiy, 2012). These methods
achieve this goal, there is often the integration of methods of analyzing data are to very large extent investigative
from the psychometrics literature and methods from the techniques and are used for estimating and projecting
machine learning and data mining literatures. An example to learning outcomes and improvement needs of institutions. In
this effect is the concurrent consideration of data generated addition, dissimilar personal behaviors can be made a
at the keystroke level, answer level, session level, student prototype so that the right steps to deal with such dissimilar
level, classroom level, and school level when examining the behaviors for the improvement of acquisition of knowledge
choice of students in the use of educational software. Issues (Corbett, 2001).
of time, sequence, and context also play important roles in
the study of educational data. With the establishment of the A solid data warehousing strategy is essential for a
annual International Conference on Educational Data successful educational data mining. The indispensability of
Mining, and the Journal of Educational Data Mining in applying relevant information in taking decisions in learning
2008,an independent area of research has ever since come to establishments has been emphasized by (Guan et al., 2002).
light. The challenge then is how to acquire quick and efficient
data to meet the needs of decision makers. There are main
Different definitions exist on EDM by various authors. factors that kick off data warehouse projects. These include
(Campbell and Oblinger, 2007) defined the analysis of data improved environment for competition, large task of passing
gathered from scholastic establishments to mean the information to stakeholders, namely community leaders,
amalgamation of methods used for mining/analyzing data in board members, parents and legislators.
the field of statistics to facilitate the promptness of staff as
well as counselors/tutors to identify learners who are likely Ideas from organizational data mining can be used in
to fail their assessment so that appropriate remedial action EDM. In mining data generated from the business
can be taken. By this, the probability of learner drop out is environments attention is centered on how establishments
reduced through the discoveries of DM. Analysis of can be assisted to maintain competitive advantage (Nemati
scholastic data centers attention on course of actions which and Barko, 2004). The reliability of ODM on organizational
took place at various levels like institution, department, theory as a field of reference is what makes it different from
school, and class. Data analysis of this sort leaves out EDM (Nemati and Barko, 2004). The benefits that
further information about every subject, and for that matter organizations can reap if they are able to transform their
it gives a general description. It comes under Educational data efficiently into valuable information and knowledge are
Data Mining. tremendous. Such benefits include augmented
competitiveness, improved decision-making, and
Evaluation of the two definitions points to a fact that prospective financial growth/increase (Nemati and Barko,
academic analysis of academic data has to do specifically 2004).
with data associated with good organization of learning
establishments as well as drop out of learners. EDM centers attention primarily on quantitative
analyses. To back case researches in EDM, interviews and
(Baker and Yacef, 2009)defines Educational Data document analysis which are qualitative techniques are
Mining as a developing field of research that generates used. Quantitative research remains the principal research
techniques capable of analyzing distinct educational paradigm of EDM with associations, predictions,
information from academic institutions and utilizing the classifications, and clusters as forms of corresponding
information to appreciate better student behavior and their results. The disadvantage of done in EDM is that outcome
academic environment. This definition left out data mining of such studies cannot be generalized for establishments
being mentioned, giving room to researchers to explore and within similar domain. What it implies is, outcomes
develop further means of analysis that can be relevant to extremely relate to an exclusive institution at a definite time.
educational data. One advantage of their research is the
broad representation of the EDM field it provides by
thrashing out issues in the outstanding research reports.
Their research however used a number of article citations as
a means of evaluating/appraising the growth of EDM.

IJISRT20MAY635 www.ijisrt.com 1924

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
C. Approaches of Mining Educational Data to use methods that are not data-mining, whether a point in
Within EDM, there exists an extensive diversity of time a student is gaming (Baker et al., 2008) by the use of
modern renowned methods. The general categories are methods of observable label to label small sets of data, a
forecasting, homogeneous / heterogeneous grouping, prediction model was developed. This was by the use of data
association discovery, using prototypes to extract collected automatically from communications between the
knowledge, and filtering raw facts. Prediction, clustering, gaming software and the students for predictor variables.
and relationship mining are widely recognized across types The problems the researchers seek to solve within the
of DM while prototypes to extract knowledge and filtering confines of whole group of raw facts and figures were made
raw facts are forms that particularly obtain eminence within possible.
EDM.
Regression, classification, and density estimation are
(Baker and Yacef, 2009; Baker, 2010) suggests four mainly the three prediction types.
key areas of application for EDM: improving student
models, improving domain models, studying the In classification, the predicted variable is a dual or
pedagogical reinforcement a program designed for tutorial discrete changeable value. Methods regarded as standard
makes available, investigations based on Science to study approaches are decision trees, logistic regression, and
students and their learning processes. Prediction, clustering, support vector machines. In regression, the forecasted
mining data to discover associations, filtering data to changeable value is a changeable value that has varying
enhance comprehension, and uncovering knowledge through quantities. Approaches for measuring the level of
the use of pattern were the five methodologies proposed by association widely employed in EDM regression that can be
them. (Castro et al., 2007) proposed the following EDM mapped onto a straight line, neural set of connections, and
subjects/tasks: program for assessing learner support vector machine regression. Regression on a straight
accomplishment, program for providing adjustment and line is the least difficult kind of regression where facts and
counsel to scholars taking into consideration their individual figures are represented on a straight line. The value that
differences in learning, methods of assessing teaching aids assesses the level of association is derived through a method
and academic programs online, programs designed to called least squares. A kind of regression that handles many
provide assessment reports in lessons delivered and received forecaster variables (Multiple Regression) is the expanded
electronically to learners as well as teachers, and designing form of linear regression and engages many changeable
programs that will identify distinctive traits of students in values. It is possible to transform nonlinear regression to
acquisition of knowledge. linear regression. An example is polynomials.

1) Forecasting The most widely used methods in determining the

Forecasting focuses on building a prototype for level of association between variables include polynomial
making inference on a distinct portion of the data referred to regression, linear regression, multi-linear interpolation,
as predicted variable, based on the integration of other robust regression, cascade correlation, radial basis functions
portion of the data known as predictor variables. In (RBFs), multivariate adaptive regression splines (MARS),
prediction it required to have labels for the variable whose group methods data handling (GMDH), and regression trees.
output is being determined for a restricted set of data. The Methods of regression were employed for forecasting the
label is a representation of information about the value of degree of learners’ comprehension based on the causes of
the output variable in distinct cases obtained from a reliable their mistakes (Feng et al., 2005). In establishing
source. In some cases, it is important however in some cases comparative proficiency of sundry methods of teaching,
to give consideration to the degree of reliability of the (Feng et al., 2009) employed breakdown method of learning,
label’s approximation. method of knowledge discovery in data obtained from
academic institutions, and methods of logistic regression for
In EDM there are two main uses of prediction. In some establishing the amount of knowledge acquisition that
cases, studying the structures of a model is essential to occurred as a result of dissimilar instructional technique
prediction and it’s a use of methods of prediction, which being applied to the impartation of similar aptitude.
gives information concerning the fundamental construct.
Research programs try to forecast the educational results of Linear correlation is a trendy approach used to
students and this is a shared approach in research programs evaluate a forecaster’s appropriateness (Bradley, 2007). In
(Romero et al, 2008). Another way by which methods of general terms, measuring precision in percentages is not a
prediction are made use of is to forecast what the resulting desired choice in categorization. This is because precision
value may be, in contexts where obtaining a label of that estimates depend extremely on initial percent on hundreds of
construct would not be desired. Considering a research that various groups. In the determination of a predictor’s
tried to examine the connection between the system of appropriateness, making sure that the various observations
gaming and learning is an example of trying to thrive in an on a particular learner are dependent is not negotiable.
environment of interactive learning by utilizing Investigators in the field of EDM employ methods that go
characteristics of the system instead of learning the material. far beyond the ordinary for analysis to make up for
Review of the structure in twelve month of during which the incomplete dependence (Strube, 1985).
learning program in various learning environment by a
researcher, may not be amenable to a direct access, putting

IJISRT20MAY635 www.ijisrt.com 1925

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
2) Predicting Student Performance Varied kinds of regulation-centered schemes were
Estimating the value yet to be determined for a employed for forecasting learner a ccomplishment in
changeable entity that explains issues about the student is settings that acquire knowledge through electronic media
the essence of prediction. The predicted values typically (Nebot et al., 2006); to forecast student accomplishment on
include accomplishment, comprehension, and grade, which the basis of learning collections built(Chen et al., 2007); for
can either be statistical and suitable for regression or forecasting, supervising and assessment of scholastic
definite, which is suitable for classification. The purpose of accomplishment of learners (Ogor, 2007); for predicting
regression is to verify the association that a dependent ratings at the end of a course of study on the premise of
changeable entity has with a single or additional characteristics mined from logged data in online educational
independent changeable entities (Draper and Smith, 1998). establishments (Shangping and Ping, 2008); for predicting
In the process of classification, entities are grouped on the learner accomplishment to offer well-timed tutorials in
basis of numerical data with regard to a singular or online e-learning environment (Chan, 2007); to forecast
additional intrinsic features of the entity on the basis of a learner’s score in web-based systems (Etchells et al., 2006).
training set of entities that were formally categorized
(Espejo et al., 2010). Several regression techniques have been used to
predict students’ marks in an open university (using model
Among the ancient and generally applied in EDM is trees, neural networks, linear regression, locally weighed
student’s accomplishment forecasting, in which many linear regression and support vector machines) (Kotsiantis
dissimilar methods as well as models are employed. and Pintelas, 2005); for forecasting responsibility evaluation
marks through linear regression projection patterns at end of
Methods of machine learning had been compared to the year (Anozie and Junker, 2006); to forecast learner
carry out forecasting of accomplishment in a program of accomplishment from logbook and quiz marks in online
study, whether good or poor in ITS (Hamalainen and Vinni, learning systems through multivariate regression pattern (Yu
2006). Further contrasting was made of various step-by-step et al., 1999); for forecasting learner scholastic
manner of mining data for the purpose of grouping learners, accomplishment by employing stepwise linear regression
and forecasting their final scores in the context of data (Golding and Donalson, 2006); employing multivariate
generated from using Moodle (Romero et al., 2008); for linear regression to forecast learning page duration (Arnold
forecasting of learner accomplishment on the premise of et al., 2005); multivariate regression used for detecting
attributes observed in user records (Minaei-bidgoli et al., estimates capable of predicting good attainment that could
2003) and for predicting scholastic accomplishment of predict success in institutions curricula (Martinez, 2001);
students in the university (Ibrahim and Rusli, 2007). regression and decision trees analysis were used to forecast
learner contentment in higher institution (Thomas and
Using models of neural network of various kinds, the Galambos, 2004); for forecasting examination results in
achievement of a learner at the end of a program has been distance learning programs (Myller et al., 2002); for
forecasted (Gedeon and Turner, 1993); for the prediction of forecasting when a student will get a question correct and
the amount of mistakes a learner will commit (Want and association rules to guide a search process to find transfer
Mitrovic, 2002); for the prediction of accomplishment on the models to predict a student’s success (Freyberger et al.,
basis of marks obtained in quiz (Girones and Fernandez, 2004); to forecast the likelihood of a learner providing the
2006); to forecast scholars’ grades, whether successful or right response to question in ITS (Cetintas et al., 2009); for
not successful on the premise of logged data from Moodle forecasting responsibility evaluation marks at year ending
(Delgado et al., 2006) and for the prediction of probable (Anozie and Junker, 2006), for predicting a learner’s quiz
accomplishment of an applicant well-thought-out for mark (Feng et al., 2005) and for predicting the likelihood of
entrance into an educational institution (Adebanjo et al., a learner’s subsequent answer to be right (Beck and Woolf,
2008). Bayesian networks were employed for the prediction 2000).
of achievement of learners considered for enrollment
(Haddawy et al., 2007); for modeling learner comprehension Conclusively, correlation analysis was employed
and forecasting of scholar accomplishment from a teaching collectively to forecast accomplishment in web-based
scheme (Pardos et al., 2007); for predicting an upcoming lessons (Wang and Newlin, 2002); to forecast a learner’s
graduate’s CGPA using candidate’s experience during final examination mark in web-based teaching (Pritchard
enrollment (Hien and Haddawy, 2007); for modeling dual and Warnakulasooriya, 2005) and to forecast the possibility
dissimilar methods to establish the possibility of correcting a of learners from secondary institutions excelling in
question demanding a lot of aptitude (Pardos et al., 2008) university (Mcdonald, 2004).
and for predicting upcoming collective accomplishment in
direct cooperative studies (Cooper et al., 2005); for 3) Relationship Mining
forecasting accomplishment at the end of a period of studies Relationship mining, seeks to achieve the objective of
in online system (Ayers and Junker, 2006); and for determining the associations between variables, considering
forecasting the result of an entity reaction (Desmarais et al., a set of data with a great number of variables. This could be
2006). by finding out the variables that associate strongly with a
variable of specific interest, or by trying to know the
strongest relationship between any two variables of interest.

IJISRT20MAY635 www.ijisrt.com 1926

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Largely, relationship mining is of four types, namely knowledge in a sphere have sequentially become the
mining to establish causal relationship (association rule), determinant for the assessments of student knowledge,
mining to detect the level of association (correlation which is by and large presented as a relation connecting
mining), mining to identify sequence of patterns, and tasks within the knowledge components and educational
fundamental mining of data. Objectively, mining to detect a program.
causal relationship is to obtain a rule of conditionality where
the presence of a group of changeable values defines in Within the instance of mining to discover
general terms a unique value of another variable. relationships, there is analysis of the relationships
concerning further changeable elements and the projections
In mining to ascertain the degree of association, of the model developed. A researcher can be enabled to
determining whether the bond is positive or negative, and examine the association involving a wide variety of
whether the relationship between variables is linear is the observable constructs and a complex latent construct.
prime purpose. In mining data to establish causal
relationship, the objective is to determine if the occurrence Frequently, employing models for discoveries
of an experience accounts for another event occurring. This influences the validity of a model’s projection to be
is achieved by examining the variation between both generalized from one context to another. Generalization
occurrences or observing the cause of one of the occurrences depends on appropriate validation.
(Scheines et al., 1994).
5) Filtering of Raw Facts and Figures
Two decisive factors that relationships ought to Filtering raw facts and figures is one of the fields of
comply with are: numerical consequence, and attraction for researchers in EDM. Individuals are capable of
interestingness. Numerical assessments tools that are giving interpretations regarding raw facts and figures, in
internationally accepted are often used for assessing some instances, if properly represented, in ways outside the
numerical significance. As a result of the numerous present capacity of absolutely automatic information
assessments that are carried, there is always the need to extraction techniques. These processes are called
regulate processes leading to the discovery of stochastic information visualization methods in educational data
relationships. Post-hoc numerical techniques or modification mining. Nevertheless, the most widely used visualizations
can be used as a regulatory method for the number of methods within educational data mining are often dissimilar
assessments carried out. By the use of this technique there is to those typically applied in further ways of viewing
high assurance that each relationship discovered was never knowledge challenges (Kay et al, 2006; Hershkovitz and
probabilistic. Using the Monte Carlo method is another Nachmias, 2008).
option for assessing the entire trend discovered to establish
its probabilistic nature. There are two main reasons for the filtering of raw
facts and figures in EDM, namely categorization and for
For the purpose of reducing the collection of recognition. Information alternatively can be filtered from
conventions, the measure of degree of association, or individuals’ recognition, to assist in building and modeling
underlying connections that the researcher has uncovered, it future forecasting samples. In such an instance, sub-sections
is essential to assess every discovery in the context of its of a data set are representation of sub-sections of the set of
relevance. A great number of noteworthy associations are data are done in text (visual format), and classification is
contained in large volumes of data. The verification of the done by coders. By and large, these labels then become the
uniqueness of the findings, their proper representation of the foundation for a predictor development. This method has a
data and the effort to reduce excessive related discoveries is proof of accelerating improvements of predictive prototypes
referred to as interestingness. Among the distinct of complicated trends, is accelerated by this proven method
determinants of interestingness are coverage, cosine, (Baker and de Carvalho, 2008).
conviction, confidence, lift, correlation, leverage, and
support. Lift and cosine have been proposed by researches to D. Application of Educational Data Mining
be uniquely applicable data from educational establishments Although, using data mining in education is a recent
(Merceron and Yacef, 2008). research field, there are many works in this area. That is
because of its potentials to educational institutes.
4) Discovery with Models
Concerning the use of a model for the discovery of (Walters, and Soyibo, 2001) conducted a study to
patterns, there is the building a model of an observable fact determine accomplishment of learners in secondary
through clustering, forecasting, or in some instances institutions on five Integrated Science progression
development of knowledge. The model that is developed proficiencies with accomplishment in connection to sex,
becomes a constituent in a further investigation, for instance rating, institute site, institute kind, kind of learner, and social
prediction or mining to discover association. and financial status. The results revealed that there was a
positive significant relationship between scholastic
In the case of prediction, in predicting a new variable, accomplishment of the learner and the institution type.
the predictions of the model developed serves as created
model’s predictions are used as variables that can be used (Hijazi and Naqvi, 2006) conducted a study on the
for prediction. The representation of the constituents of learner accomplishment by selecting a sample of 300

IJISRT20MAY635 www.ijisrt.com 1927

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
comprising 225 males and 75 females, a collection of analysis was used and the objective was to forecast the
college students. The hypothesis was stated as Learner’s marks on the basis of personal aptitude of the learner.
posture towards attending lesson, time used to learn each Bayesian Networks were employed by the researchers to
day, family earning, age of learner’s mother and learner’s obtain the finest outcome of 15% as a forecast mistake.
mother's educational status are considerably connected with
learner accomplishment. By means of simple linear (Ayesha et al. 2010) employed a method of data
regression analysis, it was discovered that elements like mining called step-by-step k-means clustering to forecast
learner’s mother's educational status and learner’s family behaviors of learners in a learners’ record comprising class
earnings were highly connected with learners’ scholastic tests, middle of term and end of term exams and projects.
accomplishment. This connected statistics would be transferred to the class
instructor for the organization of final exam. The
(Bharadwaj and Pal., 2011) undertook a study in which investigation helped the instructors to bring down the
Bayesian categorization was used to assess learner declining percentage by applying the right actions at
accomplishment. 300 learners were selected as sample of appropriate time to enhance learner accomplishment.
which 226 were males and 74 were females. A total number
of 17 characteristics were considered to ascertain which of (Moriana et al. 2006) investigated the likely impact of
them have impact on the learner accomplishment in extra-curricular activities like study-related and/or sports-
Bachelor of Computer Application program in an Indian related activities on the scholastic accomplishment of
higher institution of learning called Dr. R. M. L. Awadh learners in a Spanish secondary school. The sample selected
University. Questionnaire and institution’s database were comprised 222 learners. The learners were picked out of 12
used to gather statistics on learner’s scholastic, personal and distinct institutions. Classification of the sample was done in
social and financial characteristics. Learner’s scores were two categories based on what students engaged in when
gathered from the institution’s assessment unit. Learner’s school was not in session. The data was analyzed using
rating in senior secondary exam, residence, channel of Analysis of variance (ANOVA) to substantiate the impact of
receiving tutorial, prerequisite of mother, other addiction of activities that are non-academic on the scholastic
learner, family yearly earning and status of leaner’s family accomplishment. Observation showed that the category of
were discovered to be extremely linked with scholastic students engaged in activities that do not relate to academic
accomplishment of the learner. The study concluded that work had higher scholastic accomplishment.
scholastic accomplishment of learners does not always
depend on the learner’s own endeavor. (Baradwaj and Pal, 2011) employed data mining
technique known as categorization, and specifically decision
(Tair and El-Halees, 2012) conducted a case study in tree approach to evaluate students’ scholastic
educational data mining. Their study showed how data accomplishment. The purpose of their study was to mine
mining can be used in higher education particularly to knowledge that throws light on learner’s scholastic
improve graduate students’ performance. They used accomplishment in exam at semester ending. They used
graduate students’ data collected from the college of Science students’ data from the student database including
and Technology in Khanyounis. The data covered fifteen attendance, class test, seminar, and assignment scores. This
years period. They applied data mining techniques to study helped earlier in identifying the dropouts and students
discover knowledge. Particularly, they discovered who need special attention and allows the teacher to provide
association rules and sorted the rules. Then they used two appropriate counsel and remedy.
classification methods which are Rule Induction and Naive
Bayesian classifier to forecast the status of the graduate (Al-Radaideh et al. 2006) applied data mining
student. Also they clustered the students using K-Means techniques, particularly classification to help in improving
clustering algorithm. Finally, they used outlier detection to the quality of the higher educational system by evaluating
identify all outliers in the data. student data to study the main attributes that may affect the
student performance in courses. The extracted classification
(Kotsiantis et al., 2004) applied several step-by-step rules are based on the decision tree as a classification
methods of DM for predicting the accomplishment of method; the extracted classification rules are studied and
computer science students in distance learning course in a evaluated. It allows students to predict the final grade in a
higher institution. For every individual learner, a number of course under study.
personal characteristics e.g. Sex, age, marital status and
accomplishment qualities such as scores were used as An in-depth investigation was undertaken by
parameters of a dual classification of pass or fail. The finest employing EDM technique for assessing characteristics of
result was achieved by a Naive Bayes method which gave students in acquisition of knowledge. The research was to
precision of 74%. In addition, discovery was made that establish the usefulness of mining data in tertiary institutions
grades obtained in former institute of scholastic to enhance accomplishment of students. Students’ data was
accomplishment has much greater influence that personal collected consisting of demographic and scholastic statistics,
characteristics. and data related to the subject of study from electronic
learning system. The researcher utilized data methods of
(Pardos et al., 2006) gathered facts and figures from a data extraction (decision tree) for the discovery of numerous
web-based system of teaching in Math quiz. Regression varieties of information, including rules for association and

IJISRT20MAY635 www.ijisrt.com 1928

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
rules for classification. The researcher further categorized label is a representation of information about the value of
the students into clusters for the purpose of analyzing the output variable in distinct cases obtained from a reliable
outliers. Eventually, the researcher displayed the way by source. In some cases, it is important however in some cases
which one can take advantage of the information uncovered to give consideration to the degree of reliability of the
for enhancement of learners’ accomplishment (El-Halees, label’s approximation.
2008).
Several regression techniques have been used to
E. Visualization of Data predict students’ marks in an open university (using model
Data visualization has the purpose of visualization of trees, neural networks, linear regression, locally weighed
putting indispensable information at the center of attention linear regression and support vector machines) (Kotsiantis
and support in taking of decisions. In institutes of learning, and Pintelas, 2005); for forecasting responsibility evaluation
tutor and program supervisors are capable of analyzing how marks through linear regression projection patterns at end of
students behave toward their program of study. They are the year (Anozie and Junker, 2006); to forecast learner
also capable of using information from such analysis to accomplishment from logbook and quiz marks in online
make inference about how students learn. Visualization of learning systems through multivariate regression pattern (Yu
information and statistics have been the two methods et al., 1999); for forecasting learner scholastic
generally used for performing such tasks. accomplishment by employing stepwise linear regression
(Golding and Donalson, 2006); employing multivariate
Statistics is a scientific discipline mathematically linear regression to forecast learning page duration (Arnold
inclined. It involves the gathering, examination, reading, and et al., 2005); multivariate regression used for detecting
display of facts and figures (Freedman et al., 2007). It is estimates capable of predicting good attainment that could
comparatively simple to basically analyze and interpret facts predict success in institutions curricula (Martinez, 2001);
and a program like SPSS, used for statistical analysis of regression and decision trees analysis were used to forecast
data. The complete features of data can be presented in learner contentment in higher institution (Thomas and
abstracts and descriptions of learner behaviors when data Galambos, 2004); for forecasting examination results in
generated from academic settings are analyzed graphically distance learning programs (Myller et al., 2002); for
(Wu et al., 2002). Tutors therefore have preference for data forecasting when a student will get a question correct and
analysis on facts and figures on teaching activities which are association rules to guide a search process to find transfer
not difficult to explain (Zinn and Scheuer, 2006). models to predict a student’s success (Freyberger et al.,
2004); to forecast the likelihood of a learner providing the
F. Summary right response to question in ITS (Cetintas et al., 2009); for
Mining of Data involves the use of categories of tools forecasting responsibility evaluation marks at year ending
and techniques to uncover unrecognized samples and (Anozie and Junker, 2006), for predicting a learner’s quiz
associations found in the data (Dunham, 2003). DM mark (Feng et al., 2005) and for predicting the likelihood of
similarly is by and large a single stage in the process of a learner’s subsequent answer to be right (Beck and Woolf,
discovering information, in which institutions desire to be 2000).
well informed in taking decision using fresh data. It can
therefore be said that in taking major decisions, DM as well Conclusively, correlation analysis was employed
as discovery of information plays a crucial and inevitable collectively to forecast accomplishment in web-based
role. Considering the import of the definition and lessons (Wang and Newlin, 2002); to forecast a learner’s
explanation of data mining by the various authors, all of final examination mark in web-based teaching (Pritchard
them seemed to be dwelling on the same idea. The pivotal and Warnakulasooriya, 2005) and to forecast the possibility
issue in all definitions is the discovery of valuable of learners from secondary institutions excelling in
information from patterns, trends, and relationships university (Mcdonald, 2004).It was on this premise that the
extracted from huge or complex data sets with a primary methodology was framed.
purpose of making informed decisions that will improve
efficiency. III. METHODOLOGY

Educational data mining (also referred to as “EDM”) is It is widely accepted that Research methodology
defined as the area of scientific inquiry centered on the involves how to identify the outcome of a particular problem
development of methods for making discoveries within the of an issue under study. Researchers apply various diverse
unique kinds of data that come from educational settings, measures in tackling research problems of varying sources.
and using those methods to understand students and the Therefore, how a solution is obtained or the process of
settings which they learn better. searching for a research problem is termed Methodology
(Industrial Research Institute, 2010).
Forecasting focuses on building a prototype for
making inference on a distinct portion of the data referred to The methodology by which a research is undertaken is
as predicted variable, based on the integration of other orderly and organized and it is a discipline concerned about
portion of the data known as predictor variables. In ways a research is undertaken. Importantly, how a
prediction it required to have labels for the variable whose researcher describes the work, explains and predicts trends
output is being determined for a restricted set of data. The

IJISRT20MAY635 www.ijisrt.com 1929

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
applying scientific and logical methods to acquire Essentially, the worldwide view accepted by
information from the data is termed a research methodology. researching happenings is referred to as a research paradigm
(Creswell, 2009). Generally, research paradigm is a
It is evident that research methods, an integral portion viewpoint of study which involves major themes, methods
of the research methodology comes in different aspects. of looking for responses, and fundamental suppositions
Therefore, when research methodology is mentioned it underlying the research methods.
involves not only research methods but also considers the
reason behind the techniques applied so as to enable the As said by (Johnson and Christensen, 2005), research
researcher or anyone evaluate the decisively the outcomes of paradigm is a standpoint founded by a group of agreed
the research. The reason for a research to be carried out, the conventions, standards, ideas and practice. Simply put, the
definition of the problem, formulation of hypothesis, the way in which researchers perceive or understand growing
data gathered and the process used, the reason behind the knowledge is referred to as a paradigm. Importantly,
evaluation methods adopted and other issues of concern are research paradigms aid researchers in conducting studies
generally answered once studying a problem under research effectively.
applying research methodology.
Combining methods of researching as well as
Researchers when employing Research Methodology, philosophies in research and applying the right research
orderly identify particular questions and find solutions to paradigm and philosophy assist researchers get rid of factor
them until they arrive at a conclusion. When the research that impedes the effective implementation of techniques
done is not sequential it reduces the possibility of obtaining such as finance, restrictions among others. Three distinct
a meaningful outcome (Industrial Research Institute, 2010). terms are mentioned in reference to research paradigm they
are: positivism, interpretivism and realism research
A. Research methods philosophies (Mcnabb, 2007).
All the different processes, structures and calculations
employed by researchers in studying a research are termed Positivism Philosophy: This specifically designed on
as research methods. the extremely organized method that makes generalizing and
quantifying observations possible as well as in analyzing
Generally, its structured, technical and logical which results aided by some organized data processes. This is
consists of theoretic processes, investigational studies, generally applied in natural science making it a method
numerical methods, etc. These aid researchers in gathering objectively and analytically based. This process involves
information to enable them attain a suitable solution. natural science philosophies which include constant
Mostly, technical methods of researching require gathered philosophy, general laws with the perspective of all naturally
evidences, calculations and observing of events and occurring things (Saunders, 2003).
necessarily human judgment. Hence it is adopted if it is
proven by experimenting. Every statistics concerning the research under study
can be gathered by researchers with the aid of positivism
It is widely accepted that Research methodology is philosophy. The function of researchers in the philosophy
involves how to identify the outcome of a particular problem of research is of essence to the research study as proposed
of an issue under study. Researchers apply various diverse by (Saunders, 2003) as the researchers offer an important
measures in tackling research problems of varying sources. task of objectively analyzing the gathered information to
Simply put, the entirety of processes and techniques utilized produce positive outcomes as well as accomplish the goal
by researcher in their course of work are research methods. and objective of the research.

The main object of carrying out researches is to arrive Interpretive Philosophy: It of the belief that the
at an answer; hence the information gathered as well as the societal domain of managing and commerce is sophisticated
undetermined parameters of problem at hand should co- in formulating theories and laws for example in the natural
ordinate, to increase the possibility of achieving a solution. science. The aspect of positivism philosophy that deals with
With respect to this, there are 3 categories for grouping extensive thinking is interpretive philosophy. It is of the
methods of research: view that, there exist numerous suitable solutions to every
research issue from of a simple statistics (Johnson and
i. Methods associated with collecting information. Christensen, 2010).
ii. Techniques that relate the available information with the
undetermined factors, usually called statistical methods. Interpretive Philosophy is of a vital function in
iii. Evaluating Methods employed to analyze how accurate obtaining out comes from information gathered. With
the results are. Interpretive Philosophy researchers do not necessarily
interrelate with their surroundings but try to understand by
B. Research Paradigms making judgments and making sense out of them. An
Traditionally, approaches to research are divided into individual’s character can be influenced by difference in the
two main paradigms, namely, the Qualitative/Interpretivist standards of living, different societal and traditional settings,
and the Quantitative/positivist approach. character differences and family inclination etc. (Saunders,
2003).

IJISRT20MAY635 www.ijisrt.com 1930

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Realism Philosophy: it is as well a significant 1) Population and Sample Size
philosophy structured on human interrelation, ethics and All second year students of Mawuli SHS formed the
opinions. Realism Philosophy is concentrated on the views target population for the study. In line with the objectives of
present in a location. In realism philosophy, it is accepted the study, information bordering on the various variables of
that societies’ way of making judgments as well as their interest (students’ scores/grades in BECE, students’
conduct, is influenced by the impartial reality that exist and scores/grades for the first and second years in SHS) was
that in natural science humans are not the matter of gathered on the students. A sample size of 180 was chosen
investigation. It can therefore be concluded that realism from a population of 700 based on ±7% level of precision
philosophy reveals the way humans respond to realistic and 95% confidence level. Ten sets of data were considered;
scenarios (Johnson and Christensen, 2010). first and second year marks of students in Mathematics,
English Language, Integrated Science, Social Studies, and
C. Quantitative Research/ Positivism average scores in core subjects in BECE and SHS. The data
Quantitative in a wider perspective comprises of was gathered from the 2012to 2015-year group.
gathering data and illustrating how theories relate to
researches taking it from an objective and realistic point of 2) Sampling Procedure
view. Quantifying information can be said to be the Each of the seven programs offered in Mawuli School
objective of quantitative research. Results can be form a cluster which is homogeneous in terms of the core
generalized in quantitative research as to view a sampled subjects studied under each. A probability sampling
population as the whole and measuring ideas and views technique, precisely simple random sampling was used to
sampled.Against this backdrop, this thesis therefore adopts select records from each cluster. This sampling technique
the Quantitative/Positivism research paradigm in which the was deemed appropriate because the sampling units could be
researcher made use of quantitative research approach by identified. In other words, there is a finite number of
applying statistical analysis to numerical data students in each program and a sampling frame could thus
(students‘marks) that rsulted in various parameters which be formed. Furthermore, every student has an equal chance
were interpreted and generalized for the entire year group. of being selected for the sample, and hence the use of
This research paradigm is considered the most appropriate stochastic technique.
because the study is based on the measurement of quantity
or amount. The research process is expressed in terms of one 3) Preparing Data for Analysis
or more quantities, and hence the result of this research is a Data cleaning was carried out on the data collected.
set of numbers. This refers to the preprocessing of data in order to treat
missing values(e.g., by replacing a missing value with the
D. Research Design most commonly occurring value for that attribute, or with
A research design represents a plan, structure, and the most probable value based on statistics).
strategy of investigation conceived so as to obtain answers
to research questions and to control variance. In order to This pre-processing stage is the preparation of data
achieve the primary objective of the study, which is to before the application of data mining tool and it consists on
predict the performance of students in WASSCE, a cross a set of operations, which goal is to improve data quality and
sectional study which is a form of a Quantitative research data mining methods. This stage is really important for
was employed for this study. Mawuli School was selected as successful data mining, as real world data is "dirty", i.e., it
the area for the study. mainly consists of incomplete (e.g. missing attributes),
inconsistent (e.g. different naming) and noisy data (e.g. data
A cross sectional survey approach provides a with errors). In order to summarize the data, students’ marks
quantitative or numeric description of trends, attitudes, or in BECE and SHS were converted to averages. Only the
opinions or a population by studying a sample of that core subjects were considered in this case.
population. It also deals with comprehensive enquiry into
part of a subject with similar attributes to the entire subject 4) Data Resources
under study to obtain an understanding of the whole subject. The data resources can be classified as secondary or
primary (Dawson 2002). The empirical data for this study
To obtain the required data for the study, the was gathered from a primary source using a probability
researcher was allowed to have direct access to students’ approach. The researcher collected continuous data
academic records upon request from the Academic Affairs comprising marks obtained by students in first and second
Office of Mawuli School after having explained and assured years of their study.
the head that the records would not be used for any other
purpose apart from academic purposes. 5) Data Analysis Procedure
“Statistics is a mathematical science concerning the
Quantitative analyses was done on the main themes collection, analysis, interpretation or explanation, and
and supported with statistical presentation of actual results. presentation of data” (Freedman et al., 2007).
The major findings were also summarized in line with
objectives of the study and recommendations drawn. The data collected for the research was analyzed using
correlation analysis and linear regression analysis. SPSS v16
was used for analysis of the data.

IJISRT20MAY635 www.ijisrt.com 1931

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Descriptive statistics was used to give numerical and year scores).This means the estimation or prediction of the
graphic procedures to summarize the data in clear and unknown value of one variable (final grade) from the known
understandable way. Using the numerical approach, value of the other variable (second year scores).
statistics such as the mean and standard deviation were
computed for all the four core subjects. Also, standard 7) Simple Linear Regression Analysis
deviations for the predicted mean marks were computed. The simplest form of regression, linear regression, uses
These statistics convey information about the average. The the formula of a straight line (y = ax + b) and determines the
standard deviation was computed to explain the extent to appropriate values for a and b to predict the value of y based
which the mean has been influenced profoundly by one or upon a given value of x. The equation that describes how y
more extreme data points (referred to as an outlier). It was to is related to x is known as the regression model. Simple
show the relation that set of scores has to the mean of the linear regression was used to predict values of one variable,
sample. It was calculated as the square root of the sum of the given values of another variable. The assumption that
squared deviations from the mean divided by the number of underpinned the use of this linear regression is that the target
scores minus one. attribute is a linear function of other, mutually independent
attributes. However, the model is very flexible and can work
Other statistics computed include the coefficient of well, even if the actual dependency is only approximately
determination, coefficient of regression, minimum and the linear or the other attributes are weakly correlated.
maximum predicted values for each core subject.
In this research, simple linear regression was used to
The coefficient of determination (R-square) was used predict the marks of students in their final exam given the
to evaluate the model fit of the regression equation. That is, numerical data of their first year and second year
how good all the independent variables are at predicting the examination results. In this case the first year and second
dependent variable. The value of R-square ranges from 0.0 year marks of students were the independent variables and
to 1.0 and can be multiplied by 100 to obtain a percentage of the dependent variable was the final exam mark of students.
variance explained, and this was carried out as such in this
research. 8) Regression Equation
Suppose we have a sample of size ‘n’ and it has two
The coefficients from the regression equations sets of measures, denoted by x and y. We can predict the
represent the strength and direction of the relationship values of ‘y’ given the values of ‘x’ by using the equation,
between the independent and dependent variables. called the regression equation, Y = aX + b where the
coefficients a and b are given by
Graphical method also was used for identifying
patterns in the data. Histogram was used to show the n ∑ x y−(∑ x ) (∑ y )
b= n (∑ x2 )–( ∑ x )²
distribution of students ‘marks in the various subjects under ∑ 𝑦−𝑏 ∑ 𝑥
consideration. a=
𝑛

The interpretation of relationship between the variables For the purpose of this research x is the marks obtained
(BECE grades and SHS grades) were based on the premise by students in first year, y is the marks obtained by students
that the closer the data points come when plotted to making in second year, n is the sample size (180) and Y is the
a straight line, the higher the correlation between the two predicted mark. A prediction model was then formed for the
variables, or the stronger the relationship. If the data points four core subjects studied by students. The computation that
make a straight line going from the origin out to high x- and went into the formation of each model is presented as
y-values, then the variables are said to have a positive follows:
correlation. If the line goes from a high-value on the y-axis
down to a high-value on the x-axis, the variables have a 9) Prediction Model for Students’ Final Marks in
negative correlation. Mathematics
The y-intercept is given as 16.278 and the Slope is
Data collected on students’ academic performance in given as 0.532 in table 4.2.
all core subjects was summarized into average marks. Two
sets of such average scores, one set for first year and another The slope and the y-intercept were substituted in the
set for second year were used as the independent and following linear equation to predict students’ final score in
dependent variables respectively in the regression equation Mathematics: Y = aX + b. In this case the values of a, b, x,
to build a model that would predict students’ final and y were as follows:
mark/grade.
a = 0.532
6) Regression Analysis b = 16.278
Regression was used to fit an equation to the dataset. X =First Year Score
This data mining technique/statistical tool was used for the Y = Second Year Score
investigation of relationships between the variables. It was
to establish a causal relationship between the dependent or
outcome variable (final scores) and the predictors (second

IJISRT20MAY635 www.ijisrt.com 1932

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
To predict students’ final scores, the values of a, and b To predict students’ final scores, the values of a, and b
were substituted into the following linear equation: Y = aX were substituted into the following linear equation: Y = aX
+ b and the resulting equation is: + b. This gives the equation Second Year Score = a * First
Year Score + b and the resulting regression equation given
Second Year Score = a * First Year Score + band the by Final Year Score = 0.865 * Second Year Score + 2.445,
regression equation is given by where
Final Year Score = 0.532 * Second Year Score + 16.278, a = 0.865
where b = 2.445
a = 0.532 X = Second Year Score
b = 16.278 Y = Final Score
X = Second Year Score
Y = Final Score The regression equation Final Year Score = 0.865 *
Second Year Score + 2.445was used to project students’
The regression equation Final Year Score = 0.532 * final WASSCE scores in Integrated Science.
Second Year Score + 16.278was used to project students’
final WASSCE scores in Mathematics. 12) Prediction Model for Students’ Final Marks in Social
Studies
10) Prediction Model for Students’ Final Marks in The y-intercept is given as 24.604 and the slope is
English Language given as 0.684 in table 4.11. The slope and the y-intercept
The y-intercept is given as 30.267 and the slope is were substituted in the following linear equation to predict
given as 0.465 in table 4.5. The slope and the y-intercept students’ final score in Social Studies: Y = aX + b. in this
were substituted in the following linear equation to predict case the values of a, b, x, and y were as follows:
students’ final score in English Language: Y = aX + b. in
this case the values of a, b, x, and y were as follows: a = 0.684
a = 0.465 b = 24.604
b = 30.267 X = First Year Score
X =First Year Score Y = Second Year Score
Y = Second Year Score
To predict students’ final scores, the values of a, and b
To predict students’ final scores, the values of a, and b were substituted into the following linear equation: Y = aX
were substituted into the following linear equation: Y = aX + b. This gives the equation Second Year Score = a * First
+b Year Score + b and the resulting regression equation given
by Final Year Score = 0.684 * Second Year Score +
This gives the equation: Second Year Score = a * First 24.604, where
Year Score + b and the resulting regression equation given a = 0.684
by Final Year Score = 0.465 * Second Year Score + b = 24.604
30.267, where X = Second Year Score
a = 0.465 Y = Final Score
b = 30.267 The regression equation Final Year Score = 0.684 *
X = Second Year Score Second Year Score + 24.604was used to project students’
Y = Final Score final WASSCE scores in Social Studies.

The regression equation Final Year Score = 0.465 * The computation of the coefficient of b is explained by
Second Year Score + 30.267was used to project students’ the algorithm below:
final WASSCE scores in English Language. 1. Sum the values of x
2. Sum the values of y
11) Prediction Model for Students’ Final Marks in 3. Multiply the values of x and y and find the sum
Integrated Science 4. Multiply the result in 3 by the sample size n
The y-intercept is given as 2.445 and the slope is given 5. Multiply the result in 1 by the result in 2
as 0.865 in table 4.5. The slope and the y-intercept were 6. Find the squares of the values of x and sum them
substituted in the following linear equation to predict 7. Multiply the result in 6 by the sample size n
students’ final score in English Language: Y = aX + b. in 8. Find the square of the result in 1
this case the values of a, b, x, and y will be as follows: 9. Subtract the result in 5 from the result in 4
10. Subtract the result in 8 from the result in 6
a = 0.865 11. Divide the result in 9 by the result in 10
b = 2.445
X =First Year Score
Y = Second Year Score

IJISRT20MAY635 www.ijisrt.com 1933

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
The computation of the coefficient of a, is explained in 4. Multiply the result in 2 (sum of the values of x) by the
the following algorithm: result in 3 (sum of the values of y)
1. Sum the values of y 5. Find the squares of the values of x and sum them
2. Sum the values of x and multiply the result by b 6. Find the square of the result in 2 (sum of the values of x)
3. Subtract the result in 2 from the result in 1 7. Find the squares of the values of y and sum them
4. Divide the result in 3 by the sample size n 8. Find the squares of the result in 3 (sum of the values of
5. The symbol Y refers to the predicted value of y from a y)
given value of x from the regression equation. 9. Multiply the result in 1 by the sample size n, and subtract
the result in 4
13) Assumptions 10. Multiply the result in 5 (sum of the squares of x) by the
In order to conduct a linear regression analysis, the sample size n, and subtract the result in 6 (sum of the
following assumptions about the data were made: squares of x)
1. Linearity: It is assumed that the relationship between the 11. Find the square root of the result in 10
independent and dependent variables is linear. 12. Multiply the result in 7 (sum of the squares of y) by the
2. Normality: It is assumed that the residuals of the sample size n, and subtract the result in 8
variables are normally distributed. That is, the errors in 13. Multiply the result in 11 by the result in 12
the prediction of the value of Y (the dependent variable) 14. Divide the result in 9 by the result in 13
are distributed in a way that approaches the normal
curve. E. Summary
3. Independence: It is assumed that the errors in the Research methodology is a way to systematically solve
prediction of the value of Y are all independent of one a research problem. It may be understood as a science of
another (not correlated). studying how research is done scientifically. In it we study
4. Homoscedasticity: It is assumed that the variance around the various steps that are generally adopted by a researcher
the regression line is the same for all values of the in studying his research problem along with the logic behind
independent variables. them.

14) Correlation Analysis All the methods used by a researcher during a research
The relationship between the marks obtained by study are termed as research methods. They are essentially
students in Basic Education Certificate Examination planned, scientific and value-neutral. They include
(BECE), and the marks of students at the Senior High theoretical procedures, experimental studies, numerical
School (SHS) level was studied using correlation analysis. schemes, statistical approaches, etc. Research methods help
Correlation as a tool was used to study and measure the us collect samples data and find a solution to a problem.
extent of the relationship between the two variables.
According to (Johnson and Christensen, 2005),
15. Coefficient of Correlation research paradigm is a perspective that is based on the set of
One of the statistics used in the analysis of students’ shared assumptions, values, concepts and practices. In other
data is the coefficient of correlation, ‘r’, which measures the world, paradigm can be defined as a function of how
degree of association between the two values of related researcher thinks about the development of knowledge.
variables given in the data set. That is the degree of Research paradigm is a combination of two ideas that are
association between the marks obtained by students in related to the nature of world and the function of researcher.
BECE, and the marks of students at the SHS. It takes values It helps researcher to conduct the study in an effective
from + 1 to – 1. If two sets of data have r = +1, they are said manner. Traditionally, approaches to research are divided
to be perfectly correlated positively; if r = -1 they are said to into two main paradigms, namely, the
be perfectly correlated negatively; and if r = 0 they are Qualitative/Interpretivist and the Quantitative/positivist
uncorrelated. approach. This thesis therefore adopts the
The coefficient of correlation, ‘r’, is given by the formula Quantitative/Positivism research paradigm in which the
researcher made use of quantitative research approach by
n ∑ xy−∑ x ∑ y applying statistical analysis to numerical data
r=
√(𝑛 ∑ 𝑥 2 −∑(𝑥)2 ) (n ∑ y²−(∑ y)²) (students‘marks) thatrsultedin various parameters which
were interpreted and generalized for the entire year group.
In the formula above, n is the sample size (180), x is This research paradigm is considered the most appropriate
the marks obtained by students in BECE, y is the marks because the study is based on the measurement of quantity
obtained by students in SHS, and r is the coefficient that or amount. The research process is expressed in terms of one
determines the degree of association between students’ or more quantities, and hence the result of this research is a
performance in BECE and SHS. set of numbers.
The following algorithm explains how the coefficient A research design represents a plan, structure, and
(r) is calculated: strategy of investigation conceived so as to obtain answers
1. Multiply the values of x and y and sum them to research questions and to control variance. In order to
2. Sum the values of x achieve the primary objective of the study, which is to
3. Sum the values of y predict the performance of students in WASSCE, a cross

IJISRT20MAY635 www.ijisrt.com 1934

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
sectional study which is a form of a Quantitative research The mean of the distribution is 48.98% and the
was employed for this study. Mawuli School was selected as standard deviation is 13.73. Students’ performance in
the area for the study. Mathematics declined considerably in second year giving a
mean value of 48.98%, and a good number of students fell
In conclusion the research design and various methods below the pass mark of 44%. Also the standard deviation is
used to collect and analyze the numerical data produced the quite high, indicating the wide dispersion of individual
intended results which were discussed in the next chapter. marks from the mean.

IV. DATA ANALYSIS AND RESULTS The following tables present results of simple
regression. R Square (.311) indicates that this model
The data was analyzed and the results presented in accounts for 31% of total variation in the data; the
formulas, tables and charts. Various interpretations of the proportion of the variation that is explained by the model. R
result follow every table and chart. Each result was (0.558) is the absolute value of correlation coefficient.
examined in terms of the objectives and the research Adjusted R Square is the value adjusted for the number of
questions of the study. variables in the regression model.

A. Regression Analysis of Students’ Marks in Mathematics TABLE 4.1 MODEL SUMMARY FOR MATHEMATICS
Model Summary
Table 4.1 Model Summary for Mathematics
Model Summary
R Adjusted R Std. Error of the
Model R Square Square Estimate
1 .558 .311 .307 11.42813
a. Predictors: (Constant), First Year
Scores

From the model R indicates a moderate correlation

between the dependent and independent variable. This
means there is a moderate degree of association between the
marks obtained by students in second year and the marks
they obtained in first year. 31.1% of the variance in second
year scores of students in Mathematics is explained by the
Figure 4.2. Distribution of Marks Obtained in Mathematics independent variable (first year scores of students in
by Students in First Year Mathematics). What this means is that aside other factors
that affect performance, 31.1% of the effect is associated
The mean of the distribution is 61.46% and the with students’ performance in the first year, and this
standard deviation is 14.39. Even though the mean is percentage is very significant to be considered.
61.46%, there is a significant number of students that fell
below 61.46% (figure 4.2). The standard deviation also TABLE 4.2. COEFFICIENTS FOR MATHEMATICS
shows a wide variation of individual marks from the mean. Coefficients
Unstandardized Standardized
Coefficients Coefficients
Std.
Model B Error Beta T Sig.
1 (Constant) 16.278 3.747 4.345 .000
First Year
.532 .059 .558 8.962 .000
Scores
a. Dependent Variable: Second Year
Score

The t-statistic presented in table 4.2 is 4.345 with a

significance level less than 0.0001 (p<0.0001). The farther
away this value is from zero, the more likely it is that the
effect is statistically significant. This implies that the effect
Figure 4.3. Distribution of Marks Obtained in Mathematics of the first year scores on the second year scores is not likely
by Students in Second Year due to random chance, and that there is underlying
repeatable cause. The conclusion is that students’ second

IJISRT20MAY635 www.ijisrt.com 1935

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
year performance in Mathematics depended on their The mean mark is 59.69% and the standard deviation
performance in the first year. is 9.98. Some few students could not obtain the pass mark of
44 as shown on the chart. This represents 6.1%, and more so
TABLE 4.3. RESIDUAL STATISTICS FOR about 59% of students fell below the average.
MATHEMATICS
Residuals Statistics
Std.
Maximu Deviati
Minimum m Mean on N
Predicted Value 48.977
26.9187 65.7578 7.65554 180
8
Residual - 27.2422 11.3961
.00000 180
3.64567E1 3 7
Std. Predicted
-2.881 2.192 .000 1.000 180
Value
Std. Residual -3.190 2.384 .000 .997 180
a. Dependent Variable: second year
score
Figure 4.6. Distribution of marks obtained in English
From table 4.3, the minimum predicted value of 26.9% Language by students in second year
indicates that some students fell below the pass mark of 44.
This situation is very bad in the sense that there is too much The mean of the distribution is 58.04% and the
gap between this value and the pass mark and with the standard deviation is11.734. There is a decline in
standard deviation 7.65554, the probability of many students performance over that of first year as seen on the chart,
failing is high. This prediction reflects students’ second year given the pass mark of 44% as we see more students falling
performance which recorded 35% fail, and about 52.8% below the pass mark.
falling below the average mark of 49%.
TABLE 4.4 MODEL SUMMARY FOR ENGLISH
The maximum predicted value is 65.7%, which can be LANGUAGE
graded as C4 by the West African Examination Council Model Summary
(WAEC) scheme for West African Senior Secondary Adjusted R Std. Error of
Certificate Examination (WASSCE). The mean is predicted Model R R Square Square the Estimate
as 48.9%, which comparatively is almost the same as the
mean for the distribution of students mark in the second 1 .396 .157 .152 10.80657
year. It is however lower than the mean for the distribution
of students marks in the first year (mean = 61.5%), and it is a. Predictors: (Constant), First Year
an indication of a serious fall in performance. It can Mark
therefore be concluded that students are not performing high
in Mathematics and there is indication of further decline as From the model summary, R is 0.396, an indication of
predicted by the model. a low correlation between the dependent and independent
variable. This model accounts for about 15% of total
B. Regression Analysis of Students’ Marks in English variation in the data, (adjusted R Square is 0.152). This
Language means first year marks of students account for 15.2% of
variation in second year marks of students in English
Language. Standard error of the estimate is about 10.8.

Figure 4.5. Distribution of marks obtained in English

Language by students in first year

IJISRT20MAY635 www.ijisrt.com 1936

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
TABLE 4.5. COEFFICIENTS FOR ENGLISH
LANGUAGE C. Regression Analysis of Students’ Marks in Integrated
Coefficients Science
Standardiz
ed
Unstandardized Coefficient
Coefficients s
Std.
Model B Error Beta T Sig.
1 (Constant) 30.267 4.898 6.180 .000
Firstyear-mark .465 .081 .396 5.749 .000

a. Dependent Variable: Second

Year Mark
Form table 4.5, t-value is 6.180. The level of
significance is less than 0.0001, an indication that the effect
of students’ first year performance on their second year
performance is not due to likely random chance, and is
statistically significant.
Figure 4.8. Distribution of marks obtained in Integrated
Science in first year by students
Table 4.6Residual Statistics for English Language
Residuals Statistics The mean mark for the distribution is 67.84% and the
standard deviation is 10.75. Very few students fell below the
Std. 44% pass mark. In fact, 3.3% failed, and about 30.5% fell
Minimum Maximum Mean Deviation N actually below the average of the distribution.
Predicted
37.7116 70.7460 58.0389 4.64345 180
Value
Residual -2.48573E1 25.28843 .00000 10.77634 180

Std.
Predicted -4.378 2.737 .000 1.000 180
Value
Std.
-2.300 2.340 .000 .997 180
Residual

a. Dependent Variable: Second

Year Mark

Residuals statistics show that the predicted minimum,

maximum, and mean values are 37.3%, 70.7%, and 58.03%
respectively. The standard deviation for the predicted values
is 4.6. Comparing the predicted mean with the mean of the
distribution of students marks in both first and second year Figure 4.9 Distribution of marks obtained in Integrated
(figure 4.5 and 4.6), it can be seen that they are almost the Science in second year by students
same. This statistic reveals the reliability of the regression
model. Given the minimum predicted value of 37.3% is The average mark is 61.13% and the standard
indicative of the propensity of some students failing in deviation is 13.74.More students fell below the pass mark in
English Language. In fact, this reflects students’ comparison with the first year performance showing a
performance in second year which recorded 13.3% failure decline in second year performance over performance in
(F9), about 52.8% falling below the average mark of 58.04 first year. This has also reflected in the average score
%.The maximum predicted value can be converted to a dropping from 67 to 61 accounting for about 10.8% decline
grade B3 according to WAEC grading scheme for in performance in terms of the average score.
WASSCE. This reflects students’ second year performance
which recorded much decline. This model indicates a better
performance in English Language when compared with
Mathematics pointing to more work that needs to be done in
Mathematics to remedy the situation.

IJISRT20MAY635 www.ijisrt.com 1937

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
TABLE 4.7. MODEL SUMMARY FOR INTEGRATED Residuals Statistics
SCIENCE
Std.
Model Summary Maximu Deviati
Adjusted R Std. Error of Minimum m Mean on N
Model R R Square Square the Estimate
Predicted 61.133
35.3194 87.2260 9.30187 180
1 .677 .458 .455 10.13861 Value 3
a. Predictors: (Constant), first year Residual -
24.4899 10.1102
mark 3.36216E .00000 180
2 5
b. Dependent Variable: second year 1
mark Std. Predicted
Value -2.775 2.805 .000 1.000 180
The degree of association between the dependent and
the independent variable is moderate; the correlation Std. Residual -3.316 2.416 .000 .997 180
coefficient R is given as 0.677 in table 4.7. The model
accounts for 45.8% of total variation in the data. This a. Dependent Variable: second
implies that 45.8% the variance in second year scores is year mark
explained by students’ first year scores. This percentage is From table 4.9 the minimum predicted value is 35.3%
quite significant as it means the academic performance and the maximum predicted value is 87.2%. The grades
variable has effect than the socio-economic and assigned to these values are F9 and A1 respectively
demographic variables. according to the grading scheme of West African
Examination Council (WAEC) for West African Senior
TABLE 4.8 COEFFICIENTS FOR INTEGRATED
Secondary Certificate Examination (WASSCE). F9 is the
SCIENCE
worst grade and per this model some students are likely to
Coefficients have this grade. Some students are also likely to have the
Unstandardized Standardized best grade in Integrated Science as predicted by this model.
Coefficients Coefficients From the details of analyzed results in table 4.17 at page 74,
there is rather a massive decline in performance from the
Std.
lower grades to the higher grades. The percentage of
Model B Error Beta T Sig.
students who had grade A1 in first year dropped from 16.7%
1 (Constant) 2.445 4.840 .505 .614 to 7.8% recording 8.9 % drop. This decline is noted with
other grades also bringing the percentage of students that
First year
.865 .070 .677 12.275 .000 failed from 3.3% to 11.7%.
mark
a. Dependent Variable: second D. Regression Analysis of Students’ Scores in Social Studies
year mark

The t-value given in table 4.5 is 0.505 and the level of

significance is 0.614. This indicates that the effect of
students’ first year performance on their second year
performance is not due to likely random chance, but rather a
repeatable cause.

TABLE 4.9. RESIDUAL STATISTICS FOR

INTEGRATED SCIENCE
Residuals Statistics
Std.
Maximu Deviati
Minimum m Mean on N
Predicted 61.133
35.3194 87.2260 9.30187 180
Value 3
Residual -
24.4899 10.1102 Figure 4.11. Distribution of marks of students in Social
3.36216E .00000 180
2 5 Studies in first year
1
Std. Predicted The average score of the distribution is 69.54% and the
Value -2.775 2.805 .000 1.000 180
standard deviation is 11.109. From the chart (fig. 4.11) it is
Std. Residual observed that in spite of the high average score some
-3.316 2.416 .000 .997 180

IJISRT20MAY635 www.ijisrt.com 1938

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
students’ scores fell below the pass mark. Greater number of TABLE 4.11. COEFFICIENTS FOR SOCIAL STUDIES
students scored above 60%. Coefficients
Standardiz
Unstandardiz ed
ed Coefficien
Coefficients ts
Std.
Model B Error Beta T Sig.
1 (Const 24.60
3.494 7.042 .000
ant) 4
First
year .684 .050 .719 13.786 .000
score
a. Dependent Variable:
second year score

The t-value from table 4.11 is7.042 indicating that the

Figure 4.12. Distribution of marks of students in Social
effect of students’ first year performance on their second
Studies in second year
year performance in Social Studies is not due to likely
random chance but is due to underlying repeatable cause.
The average score for the distribution is 72.17% and
the standard deviation is 10.58.There is a very good
TABLE 4.12. RESIDUAL STATISTICS FOR SOCIAL
improvement in performance as failure rate is reduced
STUDIES
significantly. Averagely the distribution recorded about
4.7% increase in performance.
Residuals Statistics
TABLE 4.10. MODEL SUMMARY FOR SOCIAL
STUDIES
Std.
Model Summary Maximu Deviatio
Adjusted R Std. Error of Minimum m Mean n N
Model R R Square Square the Estimate
Predicted
1 .719 .516 .514 7.37468 48.5457 90.9572 72.1722 7.59924 180
Value
a. Predictors: (Constant), first year score Residual -
b. Dependent Variable: second year 2.36471E 14.98480 .00000 7.35405 180
score 1
The degree of association between the dependent and Std.
the independent variable is moderate; the correlation Predicted -3.109 2.472 .000 1.000 180
coefficient R is given as 0.719 in table 4.10. The model Value
accounts for about 52% of total variation in the data (R Std.
Square is 0.516).This means about 52% of the variance in -3.207 2.032 .000 .997 180
Residual
marks of students in Social Studies in second year is
a. Dependent Variable: second
explained by the marks students obtained in first year. This
year score
percentage is very significant in that it is an indication that
students’ past academic performance affects their future
performance far more than demographic and socio- Statistics given in table 4.12 show that 48.5% is the
economic variables. minimum predicted value, and about 91% is the maximum
predicted value. These values correspond to E8 and
A1respectively in the WAEC grading scheme for WASSCE.
It is not an indication that all students will pass in Social
Studies because the details of analyzed results in table 4.17
at page 81 shows that 2 students failed. The minimum value
also shows that some students have not gone too far away
from the failure zone and hence some more work needs to be
done in this subject area also. However, a closer look at the
results indicates massive improvement in second year
performance over the first year performance. There is a
massive move from lower grades to higher grades. In fact,

IJISRT20MAY635 www.ijisrt.com 1939

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
with the failure rate dropping from 3.3% to 1.1%, one can TABLE 5.1 DETAILS OF STUDENTS’ PERFORMANCE
confidently estimate a 0% fail in the final examination. IN MATHEMATICS.
Year One Year Two
E. Correlation Analysis of Students’ Performance at BECE Cum Cum
and SHS Gra . Perce Gra . Perce
TABLE 4.13. CORRELATION BETWEEN No. No.
de Freq nt de Freq nt
PERFORMANCE IN BECE AND SHS . .
Correlations A1 20 11.1 A1 4 2.2
BECE SHS B2 12 32 6.7 B2 6 10 3.3
B3 15 47 8.3 B3 4 14 2.2
B Pearson C4 24 71 13.3 C4 9 23 5.0
1 .518**
E Correlation C5 27 98 15.0 C5 12 35 6.7
C Sig. (2-tailed) .000 C6 25 123 13.9 C6 20 55 11.1
E D7 19 142 10.6 D7 30 85 16.7
N 180 180
E8 18 160 10.0 E8 32 117 17.8
S Pearson F9 20 180 11.1 F9 63 180 35.0
.518** 1
H Correlation Total Percent 100.0 Total Percent 100.0
S Sig. (2-tailed) .000
N 180 180 The pattern and trend of performance of students in
Mathematics extracted from their academic data are
**. Correlation is significant at the 0.01 indicative of an alarming decline which will eventually
level (2-tailed). reflect in the final examination if nothing is done about it.
Better administrative, pedagogical and timely decisions can
As indicated in table 4.13 there is a significant degree be taken based on the knowledge discovered for improved
of correlation between BECE result and SHS performance. performance in Mathematics.
The correlation coefficient is 0.518, and it means students’
performance in BECE to a large extent determines their B. Performance of Students in English Language
performance at SHS. The fact that the correlation coefficient Performance in English Language in first year was
is not exactly 1(perfect correlation), is an indication that generally good. 92.3% had a score of 50% and above while
there exist other factors that contribute to performance at the 7.7% fell below the 50% mark. This performance, however,
SHS level, most of which had been addressed in literature. declined in the second year recording 22.1% falling below
the average mark. This trend in performance can be reversed
V. DISCUSSION if teachers concentrate efforts on teaching the subject to help
improve performance.
A. Performance of Students in Mathematics
Students’ performance in Mathematics in first year TABLE 5.2. DETAILS OF STUDENTS’ PERFORMANCE
was fairly good on the average as the mean mark was 61.46. IN ENGLISH LANGUAGE.
The standard deviation however gave a clearer view of the Year One Year Two
performance. A standard deviation of 14.39 shows how the Cu Cu
individual student’s mark is widely dispersed from the Grad m. Perce Gra m. Perce
mean, indicating that there were some students who fell far No. No.
e Fre nt de Fre nt
below the average mark (21%). q. q.
A1 2 1.1 A1 3 1.6
This performance further declined in the second year B2 7 9 3.9 B2 14 17 7.8
bringing the average and the standard deviation to 48.98 and
B3 18 27 10.0 B3 14 31 7.8
13.73 respectively. This average clearly fell below 50.0 and
C4 24 51 13.3 C4 21 52 11.7
this means many more students could not obtain grade D7.
C5 41 92 22.8 C5 33 85 18.3
52.8% of the students fell below the average mark that
C6 44 136 24.4 C6 29 114 16.1
corresponds to D7 as indicated in the following table.
D7 30 166 16.8 D7 26 140 14.6
E8 3 169 1.6 E8 16 156 8.8
F9 11 180 6.1 F9 24 180 13.3
Total Percent 100.0 Total Percent 100.0

IJISRT20MAY635 www.ijisrt.com 1940

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
C. Performance of Students in Integrated Science Inferentially, students ‘performance declined in the
Averagely, performance in Integrated Science was second year significantly in Mathematics, English, and
good. In the first year, 91% scored 50% and above, leaving Integrated Science. This decline cut across all the grades;
9% which fell below the average mark. Second year saw a more students had higher grades in the first year but this
decline in performance as the number of students who fell could not be maintained or improved in the second year as
below the average increased from 9% to 19.5%.The number the number of students who scored higher grades in the first
of students who had scores above average dropped from year reduced tremendously. Significant improvement was
91% to 80.5% as shown in table 5.3. This fall has also realized in Social Studies in the second year but there
reflected in the class average dropping from 67.84% to remained a lot more effort to be exerted in order to achieve a
61.13%. 100% average and above average performance.

TABLE 5.3 DETAILS OF STUDENTS’ PERFORMANCE E. The Degree of association between students’
IN INTEGRATED SCIENCE. performance in BECE and SHS
Year One Year Two There is a significant degree of correlation between
Cu Cu BECE result and SHS performance. The correlation
Grad m. Perce Gra m. Perce coefficient 0.518 shows that students’ performance in BECE
No. No. affects their performance at SHS moderately. What this
e Fre nt de Fre nt
q. q. means is that the poor performance of students in SHS
A1 30 16.7 A1 14 7.8 cannot be blamed entirely on the grades they obtained in
B2 18 48 10.0 B2 19 33 10.6 BECE. The fact that the correlation coefficient is not exactly
B3 29 77 16.1 B3 18 51 10.0 1(perfect correlation), is an indication that there exist other
C4 30 107 16.7 C4 22 73 12.2 factors that contribute to performance at the SHS level, most
C5 33 140 18.3 C5 28 101 15.5 of which had been addressed in literature.
C6 23 163 12.8 C6 23 124 12.7
D7 8 171 4.4 D7 21 145 11.7 This finding agrees with literature; (Bharadwaj and
Pal., 2011) undertook a study in which Bayesian
E8 3 174 1.7 E8 14 159 7.8
categorization was used to assess learner accomplishment.
F9 6 180 3.3 F9 21 180 11.7
300 learners were selected as sample of which 226 were
Total Percent 100.0 Total Percent 100.0
males and 74 were females. A total number of 17
characteristics were considered to ascertain which of them
D. Performance of Students in Social Studies have impact on the learner accomplishment in Bachelor of
Performance in Social Studies in the first year was
Computer Application program in an Indian higher
averagely good, recording a class average of 69.54%.
institution of learning called Dr. R. M. L. Awadh University.
However, 6.1% fell below 50% leaving 93.9 % scoring Questionnaire and institution’s database were used to gather
above 50%. Remarkably, this performance improved in the statistics on learner’s scholastic, personal and social and
second year by recording 96.7% scoring above 50% and the
financial characteristics. Learner’s scores were gathered
remaining 3.3% below 50%, as indicated in table 5.4. More
from the institution’s assessment unit. Learner’s rating in
work need to be done to push the few students who scored senior secondary exam, residence, channel of receiving
below 50% up. tutorial, prerequisite of mother, other addiction of learner,
family yearly earning and status of leaner’s family were
TABLE 5.4 DETAILS OF STUDENTS’ PERFORMANCE
discovered to be extremely linked with scholastic
IN SOCIAL STUDIES.
accomplishment of the learner. The study concluded that
Year One Year Two scholastic accomplishment of learners does not always
Cum Cum depend on the learner’s own endeavor.
Gra . Perce Gra . Perce
No. No.
de Freq nt de Freq nt Also, (Kotsiantis et al., 2004) employed several step-
. . by-step methods of DM for predicting the accomplishment
A1 32 17.8 A1 43 23.9 of computer science students in distance learning course in a
B2 30 62 16.7 B2 33 76 18.3 higher institution. For every individual learner, a number of
B3 28 90 15.5 B3 42 118 23.3 personal characteristics e.g. sex, age, marital status and
C4 34 124 18.8 C4 26 144 14.4 accomplishment qualities such as scores were used as
C5 26 150 14.4 C5 12 156 6.7 parameters of a dual classification of pass or fail. The finest
C6 16 166 8.8 C6 15 171 8.3 result was achieved by a Naive Bayes method which gave
D7 3 169 1.6 D7 3 174 1.6 precision of 74%. In addition, discovery was made that
E8 5 174 2.8 E8 4 178 2.2 grades obtained in former institute of scholastic
F9 6 180 3.3 F9 2 180 1.1 accomplishment has much greater influence than personal
Total Percent 100.0 Total Percent 100.0 characteristics.

IJISRT20MAY635 www.ijisrt.com 1941

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Consequently, it can be concluded that performance in TABLE 5.6. PREDICTED SCORES
the BECE is a determinant of performance in SHS and by Standar
implication the CSSPS is significantly a cause of decline in Minimu Maximu d
students’ performances. Mean
m m Deviati
on
F. Degree of Correlation between Students’ First and Mathemat 48.977
Second Year Performance 26.9187 65.7578 7.6555
ics 8
English 58.038
TABLE 5.5 CORRELATION COEFFICIENTS AND T 37.7116 70.7460 4.64345
Language 9
STATISTICS Integrated 61.133
Correlation R T Statistic 35.3194 87.2260 9.30187
Science 3
Coefficient Square Social 72.172
(R) 48.5457 90.9572 7.59924
Studies 2
Mathematics 0.558 0.311 8.962
English 0.396 0.157 5.749 1) Prediction of Scores in Mathematics
Language From table 5.6, the minimum score predicted for
Integrated 0.677 0.458 12.275 Mathematics by the model is about 30%, and it is indicative
Science of students not only falling below average score but failing
Social Studies 0.719 0.516 13.786 as well. This situation is very bad in the sense that there is
too much gap between this value and the pass mark and with
The correlation coefficient in each case in the table the standard deviation approximated as 7.66; the probability
above is quite significant. The correlation coefficient of many students failing is high. This prediction reflects
measures the degree of association between students’ students’ second year performance which recorded 35.00%
performance in first and second year. The closer these fail, and about 52.80% falling below the average mark of
coefficients are to 1, the stronger the bond between the first 50%.
and second year performance. It is therefore clear from the
table that there is a close association between the two The maximum predicted value is 65.76%, which can
performances in all the subjects under study except English be graded as C4 by the West African Examination Council
Language which shows a weak association. It can therefore (WAEC) scheme for West African Senior Secondary
be concluded that first year performance has a significant Certificate Examination (WASSCE). The mean is predicted
effect on second year performance. as 48.98%, which comparatively is almost the same as the
mean for the distribution of students mark in the second
The effect of the first year performance is further year. It is however lower than the mean for the distribution
explained by the R Square. The R Square explains the of students marks in the first year (mean = 61.5%), and it is
variation in second year performance which is caused by an indication of a serious fall in performance. It can
first year performance, and in each case the percentage therefore be concluded that students are not performing high
variation is quite significant. As presented in the table, for in Mathematics and there is indication of further decline as
Mathematics R2 = 31%, for English Language R2 = 16%, for predicted by the model.
Integrated Science R2 = 46%, and for Social Studies R2 =
52%. 2) Prediction of Scores in English Language
Statistics from table 5.6 show that the predicted
The association between the first year and second year minimum, maximum, and mean values are 37.71%, 70.74%,
performances and variation in performance are also and 58.03% respectively. The standard deviation for the
validated by the t Statistic. The farther the t statistic is from predicted values is 4.64. Comparing the predicted mean with
0, the more valid the R2. The implication is that the variation the mean of the distribution of students marks in both first
in performance in first and second year is not due to likely and second year (figure 4.5 and 4.6), it can be seen that they
random chance but rather there is underlying repeatable are almost the same (59.69 and 58.04). This statistic reveals
cause. the reliability of the regression model. Given the minimum
predicted value of 37.71% is indicative of the propensity of
G. Prediction of Students’ Performance in WASSCE some students failing in English Language. In fact, this
Regression models were built to predict students’ reflects students’ performance in second year which
performance in WASSCE. The four core subject areas under recorded 13.3% failure (F9), about 52.8% falling below the
study each have a predictive model as presented in the average mark of 58.04 %.The maximum predicted value can
methodology. be converted to a grade B3 according to WAEC grading
scheme for WASSCE. This reflects students’ second year
performance which recorded much decline. This model
indicates a better performance in English Language when
compared with Mathematics pointing to more work that
needs to be done in Mathematics to remedy the situation.

IJISRT20MAY635 www.ijisrt.com 1942

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
3) Prediction of Scores in Integrated Science students passed in all eight subjects, representing 55% and
From table 5.6, the minimum predicted value is only 471 qualified for tertiary level, representing 43.5%.
35.34% and the maximum predicted value is 87.23%. The This decline is so sharp when compared to 2011 and 2012
grades assigned to these values are F9 and A1 respectively results that recorded 90.6% and 89.3% of candidates that
according to the grading scheme of West African qualified for tertiary level.
Examination Council (WAEC) for West African Senior
Secondary Certificate Examination (WASSCE). F9 is the Stakeholders always expect high performance even
worst grade and per this model some students are likely to though the stock of candidates presented for WASSCE may
have this grade. Some students are also likely to have the not be the best, and most often results of poor performance
best grade in Integrated Science as predicted by this model. hit them by surprise. Perception among stakeholders of
From the details of analyzed results in table 4.17 at page 74, education in Mawuli School is that the stock of students
there is rather a massive decline in performance from the being admitted through the Computerized School Placement
higher grades to the lower grades. The percentage of System to a very large extent is affecting performance
students who had grade A1 in first year dropped from 16.7% negatively.
to 7.8% recording 8.9 % drop. This decline is noted with
other grades also, bringing the percentage of students that With the situation at hand it is worth analyzing
failed from 3.3% to 11.7%.It is therefore certain that even students’ performance, come out with a model that predicts
though the model predicted grade A1 as the maximum, only students’ final grade to give all stakeholders a general
very few students are likely to obtain it. The model predicted overview of the final result, so that attention could be given
61.13% as the mean score and 9.30 as the standard to areas of possible flops ahead of final examination.
deviation. The rather high standard deviation shows how
widely dispersed the individual scores are from the mean. Relevant literature was reviewed covering areas such
40.60% had grades attached to scores above the mean while as similar studies conducted by various researchers, the
41.90% had grades below the grades that correspond to concept of educational data mining, and data mining
scores below the mean in the second year. techniques. The researcher also sought to investigate the
effect of the Computerized School Placement System on
4) Prediction of Scores in Social Studies Students’ performance and find the extent of it.
Statistics from table 5.6 show that 48.55% is the
minimum predicted value, and approximately 90.96% is the The target population was the second year students of
maximum predicted value. These values correspond to E8 2012 to 2015 academic year. Data for the study was
and A1 respectively in the WAEC grading scheme for collected from a secondary source; from students’ academic
WASSCE. It is an indication that all students will pass in records in Mawuli School, and a sample size of 180 was
Social Studies. This, however, does not call for complacency considered for the study. Three sets of data were collected
because scores of some students definitely will fall below namely: marks of students in all core subjects in first year,
the 50% average. The minimum value also shows that some marks of students in all core subjects in second year, and
students have not gone too far away from the failure zone grades of students in all core subjects in BECE.
and hence some more work needs to be done in this subject
area also. However, a closer look at the results indicates The data was analyzed using SPSS v16. Simple linear
massive improvement in second year performance over the regression analysis was performed to analyze performance
first year performance. There is a significant move from and construct prediction models to estimate students’ final
lower grades to higher grades. In fact, with the failure rate scores in all core subjects. Also, to investigate the
dropping from 3.3% to 1.1%, one can confidently estimate a relationship and determine the degree of association between
0% failure rate in the final examination even as the model students’ performance in SHS and their grades at BECE, a
projected. The mean value predicted is 72.17% and this is Pearson’s correlation analysis was used.
consistent with the mean of students’ scores in second year
which proves the reliability of the model. Results of the study were presented in tables, charts,
and formulas. It was revealed that there was a decline in
VI. SUMMARY,CONCLUSION,AND students’ performance in Mathematics, English Language,
RECOMMENDATION and Integrated Science, and a good number of students fell
below the pass mark, although some few students were
A. Summary performing high, whiles Social Studies saw a good
Mawuli School is one of the renowned second cycle improvement in students’ performance over performance in
institutions in Ghana, and most especially in the Volta the first year by recording a 98.9% pass.
Region of Ghana. In fact, it is referred to as “Volta
University” due to its unparalleled performance in the past. It was also found that there is a close association
In recent times, however, performance does not seem to be between students’ first year and second year performance,
what it used to be, even though better than those of some of and that the effect of students’ first year performance on
its counterparts. Comparative analysis of the West African their second year performance was not due to likely random
Senior School Certificate Examination of 2013 revealed that chance but that the effect had underlying repeatable cause,
Mawuli School placed fourth position in the Volta Region. and hence models were constructed to predict students’ final
Out of 1089 candidates who took the examination, 599 WASSCE scores/grades. For Mathematics, the minimum

IJISRT20MAY635 www.ijisrt.com 1943

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
mark predicted is 26.9%, which corresponds to grade F9, in scores. This is in line with works done by earlier
and the maximum mark predicted is 65.6%, corresponding researchers. (Ayesha et al. 2010) employed a method of data
to grade C4. The average is 48.0%, and this corresponds to mining called step-by-step k-means clustering to forecast
grade E8. The minimum mark predicted for English behaviors of learners in a learners’ record comprising class
Language is 37.7, which is grade F9; the maximum mark tests, middle of term and end of term exams and projects.
predicted is 70.7%, which is grade B3, and the average is This connected statistics would be transferred to the class
58.0, which is grade C6. For Science, 35.3%, 87.2%, and instructor for the organization of final exam. The
61.1%, corresponding to grades F9, A1, C5 were predicted investigation helped the instructors to bring down the
for the minimum, maximum, and the average respectively. declining percentage by applying the right actions at
Predictions for Social Studies were 48.5%, 90.6%, and appropriate time to enhance learner accomplishment.
72.2% as minimum, maximum, and the average mark
respectively. These marks represent E8, A1, and B3 Also, (Kotsiantis et al., 2004) applied several DM
respectively. Correlation does exist between the algorithms to predict the performance of computer science
performance in BECE and performance at SHS. The students, and concluded that past school grades have a much
correlation coefficient is 0.518, and it means students’ higher impact than demographic variables.
performance in BECE to some extent accounts for their
performance at SHS. The implication therefore is that the C. Recommendations
Computerized School Placement System accounts for the Based on the findings of the study, it is recommended
poor performance of students as perceived by stakeholders. that:
The fact, however, is that because the correlation coefficient 1. Management and teachers in Mawuli School must find
is not exactly 1(perfect correlation), poor performance ways to intensify efforts to improve performance in the
cannot be blamed entirely on the Computerized School three core subject areas namely Mathematics, Integrated
Placement System, implying there are other factors that Science, and English Language.
contribute to students ‘poor performance at the SHS level. 2. Stakeholders should not have baseless expectations of
students’ performance but rather consider students’
B. Conclusion terminal performance as a guide.
The research questions stated for the study were fully 3. Management, teachers, and other stakeholders of Mawuli
examined in relation to the results of the study. It was found School must find ways to deal with other factors that
that few students performed high in Social Studies, affect students’ performance such as demographic and
Integrated Science, and English Language than in socio- economic variables like gender, grade level,
Mathematics even though performances in all core subjects school location, school type, student type, student’s
except Social Studies kept declining. Also, it was attitude towards attendance in class, hours spent in study
established from the results that there is a significant degree on daily basis after college, students' family income,
of association between performance in BECE and SHS. students' mother's age and mother's education as revealed
Performances at the two levels were moderately correlated, in literature.
meaning indeed the grades of students in BECE affect their 4. Management must initiate remedial efforts to help weak
performance at SHS. This finding agrees with literature; that students immediately they are admitted for their
the factors like students’ grade in senior secondary exam, programs in order to help improve their performance far
living location, medium of teaching, mother’s qualification, ahead of the WASSCE examination.
students other habit, family annual income and student’s
family status were highly correlated with the student VII. ACKNOWLEGMENT
academic performance (Bharadwaj and Pal, 2011). Also, it
was found that past school grades have a much higher I would first like to thank the almighty God of heaven
impact than demographic characteristics (Kotsiantis et al., for His love, goodness, support, and wisdom He provided
2004). me during the entire period of my studies in the university.
In fact, I owe the very foundation of my life and success to
However, one cannot blame poor performance on the Him.
Computerized School Placement System entirely since the
correlation coefficient does not show perfect correlation. My gratitude and respect go to Mr. Dominic Asamoah
and Dr. Ussif Najim who through patience and professional
Finally, it was clearly seen that there exists significant expertise supervised this work. I do appreciate you for your
degree of association between students’ first year scores and careful attention to each of the chapters and all the
second year scores. This implies that students’ performance suggestions and directions you offered me throughout this
in the second year of their program was affected by their work.
performance in the first year, and clearly this effect was not
due to likely random chance but had underlying repeatable Special thanks to Rev. S.S. Asieni, the Assistant
cause. This also implies that students’ terminal performance Headmaster Academic, Mawuli School, for his immense
affects their final results to a large extent. Based on this fact support toward the success of this work.
prediction models were constructed to project students’ final
scores. However, coefficients of determination in all cases Finally, kudos to all friends and course mates who
indicate that there are other factors that explain the variance shared their rich experiences, ideas, and suggestions to bring

IJISRT20MAY635 www.ijisrt.com 1944

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
this work to accomplishment. All your contributions are [13] Bharadwaj, B.K., and Pal, S. (2011). “Data Mining:
very much appreciated. A Prediction for Performance Improvement Using
Classification”. International Journal of Computer
REFERENCES Science and Information Security (IJCSIS), 9 (4), 136-
140.
[1] Adebanjo, A.T., Oladokun, V.O., and Charles-owaba, [14] Bradley, A.P. (2007). “The Use of the Area under the
O.E. (2008). ”Predicting Student’s Academic ROC Curve in the Evaluation of Machine Learning
Performance Using Artificial Neural Network: A Case Algorithms”. Pattern Recognition, 30, 1145-1159.
Study of an Engineering Course”. Pacific Journal of [15] Calders, T., and Pechenizkiy, M. (2012), “Introduction
Science and Technology, 9 (1), 72-79. to the Special Section on Educational Data Mining”.
[2] Al-Radaideh, Q., Al-Shawakfa, E., and Al-Najjar, M., SIGKDD Explorations, 13 (2) pp.3-6.
(2006). ”Mining Student Data Using Decision Trees”. [16] Campbell, J.P., and Oblinger, D.G. (2007).”Academic
The 2006 International Arab Conference on Analytics”. EDUCAUSE, Washington, D.C.
Information Technology (ACIT’ 2006). [17] Castro, F., Vellido, A., Nebot, A., and Mugica, F.
[3] Anozie, N., and Junker, B.W. (2006).”Predicting End- (2007).”Applying Data Mining Techniques to e-
of-year Accountability Assessment Scores from Learning Problems”. In Jain, L.C., Tedman, R. and
Monthly Student Records in an Online Tutoring Tedman, D. (eds.), Evolution of Teaching and
System”. Educational Data Mining AAAI Workshop Learning Paradigms in Intelligent Environment:
(pp. 1-6). California. Studies in Computational Intelligence, 62, 183-221.
[4] Arnold, A., Scheines, R., Beck, J.E., and Jerome, B. Springer-Verlag.
(2005). Time and Attention”: Students, Sessions, and [18] Cetintas, A., Si, L., Xin, Y.P., and Hord, C.
Tasks. AAAI2005 Workshop on Educational Data (2009).”Predicting Correctness of Problem Solving
Mining (pp. 62-66). Pittsburgh. from Low-level Log Data in Intelligent Tutoring
[5] Ayesha, S., Mustafa, T., Sattar, A.R., and Khan, M.I. Systems”. International Conference on Educational
(2010).”Data Mining Model for Higher Education Data Mining, (pp 230-238). Cordoba, Spain.
System. European Journal of Scientific Research”, [19] Chakrabarti, S., Thomas P., Nadeau, et al.
43(1), 24-29. (2009).”Data Mining”. Morgan Kaufmann Publishers,
[6] Ayers E., and Junker B.W. (2006). “Do Skills Burlington.
Combine Additively to Predict Task Difficulty in [20] Chadha A. and Kumar V (2011). “An Empirical Study
Eighth Grade Mathematics?” AAAI Workshop on of the Applications of Data Mining Techniques in
Educational Data Mining, pp. 14-20. Menlo Park. Higher Education”. International Journal of Advanced
[7] Baker, R. (2010). “Data Mining for Education”. In: Computer Science and Applications, 2(3), pp. 80-84.
McGraw, B., Peterson, P., Baker, E. (eds.) [21] Chan, C.C. (2007). “A Framework for Assessing
International Encyclopedia of Education, 3rd ed. Usage of Web-Based e-Learning Systems”.
Elsevier, Oxford. International Conference on innovative Computing,
[8] Baker, R. and Carvalho, A. (2008). “Labeling Student Information and Control, Washington, DC, 147- 151.
Behavior Faster and More Precisely with [22] Chen, C., Chen, M., and Li, Y. (2007). “Mining Key
TextReplays”. 1st International Conference on Formative Assessment Rules Based on Learner
Educational Data Mining, pp 38-47. Profiles for Web-based Learning Systems”. IEEE
[9] Baker, R., Barnes, T., and Beck, J.E. (2008). International Conference on Advanced Learning
“Educational Data Mining 200”8: 1st International Technologies, Japan, pp 1-5.
Conference on Educational Data Mining, Proceedings. [23] Cooper, M.,Stevens, R., Giordani, A., Soller, A and
Montreal, Quebec, Canada. Gerosa, L., (2005). “Developing a Framework for
[10] Baker, R., and Yacef, K. (2009). “The State of Integrating Prior Problem Solving and Knowledge
Educational Data Mining in 2009: A Review and Sharing Histories of a Group to Predict Future
Future Visions”. Journal of Educational Data Mining, Group Performance”. International Conference on
1 (1), 3-17. Collaborative Computing: Networking, Applications
[11] Beck, J.E., and Woolf, B.P. (2000).”High-level and Worksharing, (pp. 1-9). Boston.
Student Modeling with Machine Learning”. Fifth [24] Corbett, A. T. (2001). “Cognitive Computer Tutors:
International Conference on Intelligent Tutoring Solving the Two-Sigma Problem”. 8th International
Systems, (pp. 584-593). Alagoas, Brazil. Conference on User Modeling.
[12] Berson, A., Smith, S., and Thearling, K. (2011). “An [25] Corbett, A.T., and Anderson, J.R., (1995).
Overview of Data Mining Techniques”. Retrieved “Knowledge Tracing: Modeling the Acquisition of
from Procedural Knowledge, User Modeling and User-
http://www.thearling.com/text/dmtechniques/dmtechni Adapted Interaction, 4, 253-278.
ques.htm

IJISRT20MAY635 www.ijisrt.com 1945

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[26] Creswell, J.W (2003). “Research design: qualitative & [40] Garcia, P., Amandi, A., Schiaffino, S., and Campo, M.
quantitative approaches”. Thousand Oaks, CA: Sage (2008). “Evaluating Bayesian Networks’ Precision for
[27] Dawson, C. (2002). “Practical Research Methods: a Detecting Student’s Learning Styles”. Computer and
User-friendly Guide to Mastering Research Education Journal, 49, 794-808.
Techniques and Projects”. How to Books Ltd. Oxford [41] Gedeon, T.D., and Turner, H.S. (1993). “Explaining
OX4 IRE, UK. Student Grades Predicted by a Neural Network”.
[28] Delgado, M., Gibaja, E., Pegalajar, M.C., and Pérez, International conference on Neural Networks, (pp.
O. (2006). “Predicting Students' Marks from. Moodle 609-612). Nagoya.
Logs using Neural Network Models”. International [42] Girones, M., and Fernandez, T.A. (2006). “Ariadne, a
Conference on Current Developments in Technology- Guiding Thread in the Learning Process’s Labyrinth”.
Assisted Education, Sevilla, Spain, 586-590. International Conference on Current Developments in
[29] Desmarais, M.C., Gagnon, M., and Meshkinfram, P. Technology-Assisted Education, (pp. 287-
(2006). “Bayesian Student Models Based on Item to 290).Sevilla.
Item Knowledge Structures”. Conference on [43] Golding, P., and Donalson, O. (2006). “Predicting
Technology Enhanced Learning, (pp 1-10). Crete, Academic Performance”. Frontiers in Education
Greece. Conference, (pp 21-26). San Diego, California.
[30] Draper, N.R., and Smith, H., (1998).”Applied [44] Guan, J., Nunez, W., and Welsh, J. (2002).
Regression Analysis”.Wiley. “Institutional Strategy and Information Support”: The
[31] Dunham, M. (2003). Data Mining: “Introductory and Role of Data Warehousing in Higher Education.
Advanced Topics”. Pearson Education, Upper Saddle Campus-Wide Information Systems, 19 (5), 168-174
River, NJ. [45] Haddawy, P., Thi, N., and Hien, T.N. (2007). “A
[32] El-Halees, A. (2008). “Mining Students Data to Decision Support System for Evaluating International
Analyze Learning Behavior: A Case Study”. The 2008 Student Applications”. Frontiers in Education
international Arab Conference of Information Conference, (pp. 1-4). Milwaukee.
Technology (ACIT2008) – Conference Proceedings. [46] Hamalainen, W., and Vinni, M. (2006). “Comparison
University of Sfax, Tunisia, Dec 15- 18. of Machine Learning Methods for Intelligent Tutoring
[33] Espejo, P., Ventura, S., and Herrera, F. (2010).”A Systems”. International conference in intelligent
Survey on the Application of Genetic Programming to tutoring systems, (pp. 525-534). Taiwan.
Classification”. IEEE Transactions on Systems, Man, [47] Han, J. and Kamber, M. (2001). “Data Mining:
and Cybernetics-Part C, 40, (2), 121-144. Concepts and Techniques”. Simon Fraser University,
[34] Etchells, T.A., Nebot, A., Vellido, A., Lisboa, P.J.G., Morgan Kaufmann Publishers, ISBN 1-55860-489-8.
and Mugica, F. (2006). “Learning what is Important: [48] Hershkovitz, A., and Nachmias, R. (2008).
Feature Selection and Rule Extraction in a Virtual “Developing a Log-based Motivation Measuring
Course”. European Symposium on Artificial Neural Tool”. 1st International Conference on Educational
Networks, Bruseles, Belgium pp 401-406. Data Mining, (pp. 226-233). Montreal.
[35] Fayyad, U., Piatetsky-Shapiro, G., and Smyth, P. [49] Hijazi, S. T., and Naqvi, R. S. M. M. (2006). “Factors
(1996). “From Data Mining to Knowledge Discovery: Affecting Student’s Performance: A Case of Private
An Overview”. Advances in Knowledge Discovery Colleges”. Bangladesh e-Journal of Sociology, 3, (1).
and Data Mining, MIT Press, pp 1-3. (pp. 97-106).
[36] Feng, M., Beck, J., and Heffernan, N. (2009).”Using [50] IBM. (2012). “What is big data?” Retrieved from
Learning Decomposition and Bootstrapping with 01.ibm.com/software/data/bigdata/
Randomization to Compare the Impact of Different [51] Ibrahim, Z., and Rusli, D. (2007). “Predicting
Educational Interventions on Learning”.2nd Students’ Academic Performance: Comparing
International Conference on Educational Data Mining, Artificial Neural Network, Decision Tree and Linear
(pp 51-60).Cordoba, Spain. Regression”. Annual SAS Malaysia Forum, (pp. 1-6).
[37] Feng, M., Heffernan, N., and Koedinger, K. (2005). Kuala Lumpur.
“Looking for Sources of Error in Predicting Student’s [52] Industrial Research Institute (2010). “Research
Knowledge”. AAAI’05 workshop on Educational Management”. Industrial Research Institute, Michigan.
Data Mining, (pp 1-8). Amsterdam. [53] Johnson, B., and Christensen, L., (2010). “Educational
[38] Freedman, D., Purves, R., and Pisani, R. (2007). Research: Quantitative, Qualitative, and Mixed
“Statistics”. 4th Edition. W.W. Norton & Co. Approaches”. UK: SAGE.
[39] Freyberger, J., Heffernan, N.T., and Ruiz, C. (2004). [54] Kay, J., Maisonneuve, N., Yacef, K., and Zaiane, O.R.
Using Association Rules to Guide a Search for Best (2006). “Mining Patterns of Events in Students'
Fitting Transfer Models of Student Learning”. Teamwork Data”. Proceedings of Educational Data
Workshop Analyzing Student-Tutor Interaction Logs Mining Workshop, (pp 1-8). Taiwan.
to Improve Educational Outcomes, (pp.1-10). Alagoas,
Brazil.

IJISRT20MAY635 www.ijisrt.com 1946

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[55] Kiron, D., Shockley, R., Kruschwitz, N., Finch, G., [69] Pardos Z.; Heffernan N.; Anderson B.; and Heffernan
and Haydock, M. (2012). Analytics: “The Widening C., (2006). “Using Fine-Grained Skill Models to Fit
Divide”. MIT Sloan Management Review, 53(2), 1-22 Student Performance with Bayesian Networks”.
[56] Kotsiantis, S.B., and Pintelas, P.E., (2005). “Predicting Proceedings of 8th International Conference on
Students' Marks in Hellenic Open University”. IEEE Intelligent Tutoring Systems. Taiwan.
international Conference on Advanced Learning [70] Pardos, Z., Heffernan, N., Anderson, B., and
Technologies, (pp. 664-668). Washington, DC. Heffernan, C. (2007). “The Effect of Model
[57] Kotsiantis, S., Pierrakeas, C., and Pintelas, P., (2004). Granularity on Student Performance Prediction Using
“Prediction of Student’s Performance in Distance Bayesian Networks”. International Conference on User
Learning Using Machine Learning Techniques”. Modeling, (pp. 435-439). Corfu, Greece.
Applied Artificial Intelligence, Vol. 18, No. 5, pp. 411- [71] Pardos, Z., Beck, J.E., Ruiz, C., and Heffernan, N.
426. (2008). “The Composition Effect: Conjunctive or
[58] Larose, D., (2006). “Data Mining: Methods and Compensatory? An Analysis of Multi-Skill Math
Models”. John Wiley. Questions in ITS”. International Conference on
[59] Malik, K., and Ranjan, J., (2007). “Effective Educational Data Mining, (pp. 147-156). Montreal.
Educational Process: A Data Mining Approach”. [72] Pritchard, D., and Warnakulasooriya, R. (2005). “Data
VINE: The Journal of Information and Knowledge from a Web-based Homework Tutor Can Predict
Management Systems, 37(4), 502-515. Student’s Final Exam Score”. World Conference on
[60] Martinez, D., (2001). “Predicting Student Outcomes Educational Multimedia, Hypermedia and
Using Discriminant Function Analysis”. Meeting of Telecommunications, (pp. 2523-2529). Chesapeake.
the Research and Planning Group, pp. 1-22. Lake [73] Rajagopalan, S. P., and Shyamala, K., (2006). “Data
Arrowhead, CA. Mining Model for a Better Higher Educational
[61] Mcdonald, B., (2004). “Predicting students’ success”. System”. Information Technology Journal, 5 (3), 560-
Journal for Mathematics Teaching and Learning, (pp. 564.
1-14). [74] Romero, C. and Ventura, S. (2007). “Educational Data
[62] Merceron, A., and Yacef, K. (2008). “Interestingness Mining: A survey from 1995 to 2005”. Expert
Measures for Association Rules in Educational Data”. Systems with Applications, 33, 135-146.
International Conference on Educational Data Mining, [75] Romero, C., Ventura, S., Hervás, C., and Gonzales, P.
(pp. 57-66) Montreal, Canada. (2008). “Data Mining Algorithms to Classify
[63] Minaei-bidgoli, B., Kashy, D.A., Kortmeyer, G., and Students”. International Conference on Educational
Punch, W.F. (2003). “Predicting Students’ Data Mining, (pp. 8-17). Montreal, Canada.
Performance: An Application of Data Mining Methods [76] Sargenti, P., Lightfoot, W. and Kehal, M. (2006).
with an Educational Web-based system”. International “Diffusion of Knowledge in and through Higher
Conference on Frontiers in Education, (pp.13-18). Education Organizations”. Issues in Information
Colorado, USA. Systems, 3 (2), 3-8.
[64] Moriana, J.A., Alos, F., Alcala, R. Pino, M. J., [77] Saunders, M., (2003). “Research Methods for Business
Herruzo, J., et al. (2006). “Extra-Curricular Activities Students”. Pearson Education, South Africa.
and Academic Performance in Secondary Students”. [78] Scheines, R., Sprites, P., Glymour, C., and Meek, C.
Electronic Journal of Research in Educational (1994). “Tetrad II: Tools for Discovery”. Lawrence
Psychology, 4 (1), 35-46. Erlbaum Associates: Hillsdale, NJ.
[65] Myller, N., Suhonen, J., and Sutinen, E. (2002). [79] Shangping, D., and Ping, Z. (2008). “A Data Mining
“Using Data Mining for Improving Web-Based Course Algorithm in Distance Learning”. International
Design”. International Conference on Computers in Conference on Computer Supported Cooperative Work
Education, (pp. 959- 964). Washington. in Design, (pp. 1014-1017), Xian.
[66] Nebot, A., Castro, F., Vellido, A., and Mugica, F. [80] Strube, M.J. (1985). “Combining and Comparing
(2006). “Identification of Fuzzy Models to Predict Significance Levels from Non Independent Hypothesis
Students’ Performance in an e-Learning Environment”. Tests”. Psychological Bulletin, 97, 334-341.
International Conference on Web-based Education, [81] Tair, M.M.A., and El-Halees, A.M., (2012). “Mining
(pp. 74-79). Puerto Vallarta. Educational Data to Improve Students’ Performance”.
[67] Nemati, H., and Barko, C. (2004). “Organizational International Journal of Information Technology
Data Mining (ODM): An Introduction”. In H. Nemati Research, 2(2).
and C. Barko (Eds.), Organizational Data Mining (pp. [82] Thomas, E.H., and Galambos, N. (2004). “What
1-8). London: Idea Group Publishing. Satisfies Students? Mining Student-Opinion Data with
[68] Ogor, E.N. (2007). “Student Academic Performance Regression and Decision Tree Analysis”. Research in
Monitoring and Evaluation Using Data Mining Higher Education Journal, 45 (3), 251-269.
Techniques”. Electronics, Robotics and Automotive
Mechanics Conference, (pp. 354-359). Washington,
DC.

IJISRT20MAY635 www.ijisrt.com 1947

Volume 5, Issue 5, May – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[83] Walonoski, J., and Heffernan, N.T. (2006). “Detection
and Analysis of Off-Task Gaming Behavior in
Intelligent Tutoring Systems”. Proceedings of the 8th
International Conference on Intelligent Tutoring
Systems, 382-391.
[84] Walters, Y.B., and Soyibo, K. (2001). “An Analysis of
High School Students' Performance on Five Integrated
Science Process Skills”. Research in Science
&Technical Education, 19 (2), 33 – 145.
[85] Wang, A.Y., and Newlin, M.H. (2002). Predictors of
Web-Based Performance: The Role of Self-Efficacy
and Reasons for Taking an On-Line Class”. Computers
in Human Behavior Journal, 18, 151-163.
[86] Want, T., and Mitrovic, A. (2002). “Using Neural
Networks to Predict Student's Performance”.
International Conference on Computers in Education,
(pp. 1-5). Washington, DC.
[87] Witten, I.H. and Frank, E. (1999). “Data Mining:
Practical Machine Learning Tools and Techniques
with Java Implementations”. Morgan Kauffman, San
Francisco, CA.
[88] Wu, A., and Leung, C. (2002). “Evaluating learning
behavior of Web-Based Training (WBT) using Web
log”. International Conference on Computers in
Education, (pp. 736-737).New Zealand.
[89] Yu, C.H., Jannasch-pennell, A., Digangi, S., and
Wasson, B. (1999). “Using On-line Interactive
Statistics for Evaluating Web-based Instruction”.
Journal of Educational Media International, 35,157-
161.
[90] Zinn, C., and Scheuer, O. (2006). Getting to Know
Your Students in Distance Learning Contexts”.1st
European Conference on Technology Enhanced
Learning, (pp. 437-451).

IJISRT20MAY635 www.ijisrt.com 1948