STUDENT PERFORMANCE ANALYSIS AND PREDICTION

USING MACHINE LEARNING

A Project Report Submitted to

JNTUA, Ananthapuramu

In partial fulfillment of the requirements for the award of the degree of

Bachelor of Technology

INFORMATION TECHNOLOGY
By
D. ROSHINI (20KB1A1212) M. VENKATA KALYAN (20KB1A1236)
S. GEETHIKA PRIYA(20KB1A1251) T. JAYANTH(20KB1A1252)

Under the esteemed Guidance of

Dr. A. NARAYANA RAO

PROFESSOR, Dept. of IT and AI&DS

DEPARTMENT OF IT and AI&DS

N.B.K.R INSTITUTE OF SCIENCE & TECHNOLOGY
(Autonomous)

VIDYANAGAR – 524 413, TIRUPATI DIST, AP

MAY 2024
Website: www.nbkrist.org. Ph: 08624-228 247

Email: [email protected]. Fax: 08624-228 257

N.B.K.R. INSTITUTE OF SCIENCE & TECHNOLOGY

(Autonomous)
(Approved by AICTE: Accredited by NBA: Affiliated to JNTUA, Ananthapuramu)
An ISO 9001-2000 Certified Institution

Vidyanagar -524 413, Tirupati District, Andhra Pradesh, India

BONAFIDE CERTIFICATE

This is to certify that the project work entitled “STUDENT PERFORMANCE

ANALYSIS AND PREDICTION USING MACHINE LEARNING” is a bonafide

work done by D.ROSHINI (20KB1A1212), M.VENKATA KALYAN (20KB1A1236),

S.GEETHIKA PRIYA (20KB1A1251), T.JAYANTH (20KB1A1252), in the department of IT

and AI&DS, N.B.K.R. Institute of Science &Technology, Vidyanagar and is
submitted JNTUA, Ananthapuramu in the partial fulfilment for the award of B.Tech
degree in Information Technology. This work has been carried out under our supervision.

Dr. A. NARAYANA RAO

Professor & HOD
Department of IT AND AI&DS
N.B.K.R.I.S.T

Submitted for the Viva-Voce Examination held on

Internal Examiner External Examiner

 ACKNOWLEDGEMENT

The satisfaction that accompanies the successful completion of a project would

be incomplete without the people who made it possible of their constant guidance and
encouragement crowned our efforts with success.
We would like to express our profound sense of gratitude to our project guide Dr.
A. Narayana Rao, Professor, Department of IT and AI&DS, N.B.K.R.I.S.T (affiliated to
JNTUA, Ananthapuramu), Vidyanagar, for his masterful guidance and the constant
encouragement throughout the project. Our sincere appreciations for his suggestions and
unmatched services without, which this work would have been an unfulfilled dream.

We convey our special thanks to Dr. Y. Venkata Rami Reddy

respectable chairman of N.B.K.R. Institute of Science and Technology, for
providing excellent infrastructure in our campus for the completion of the project

We convey our special thanks to Sri N. Ram Kumar Reddy respectable

correspondent of N.B.K.R. Institute of Science and Technology, for providing excellent
infrastructure in our campus for the completion of the project.

We are grateful to Dr. V. Vijaya Kumar Reddy, Director, of N.B.K.R Institute

of Science and Technology for allowing us to avail all the facilities in the college.

We would like to convey our heartful thanks to Staff members, Lab

technicians, and our friends, who extended their cooperation in making this project as a
successful one.

We would like to thank one and all who have helped us directly and
indirectly to complete this project successfully.

i
 Table of Contents

Chapter Topic Name Page

No No

Acknowledgement i

List of Tables iv

List of Figures V

Abstract vi

1 Introduction 1-7

1.1 Introduction 2-3

1.2 Background and Motivation 3-4

1.3 Problem Statement 4

1.4 Objectives and Scope 5

1.5 Organization of the project Report 6

1.6 Summary 7

2 Literature Review 8-17

2.1 Introduction 9

2.2 Literature Survey 10-13

2.3 Identification of Research Gap 14-16

2.4 Summary 17

3 Methodology 18-26

3.1 Introduction 19

3.2 Overview of Methodological Approach 19-20

3.3 Parameters 21-22

3.3 Performance Metrics 23

3.4 Description of Tools and Technologies Used 24-25

3.5 Summary 26

ii
 Table of Contents

4 System Design 27-34

4.1 Introduction 28

4.2 Detailed Design of Components 29-32

4.3 UML Diagrams 33-34

4.4 Summary 35

5 Implementation 36-44

5.1 Introduction 37

5.2 Code Structure and Organization procedure 38-39

5.3 Algorithms and Techniques Implemented 40-43

5.4 Summary 44

6 Results and Analysis 45-57

6.1 Introduction 46

6.2 Results 47-53

6.3 Analysis of Performance Metrics 54

6.4 Discussion of Findings 55-56

6.5 Summary 57

7 Conclusion and Future Work 58-61

7.1 Conclusions Drawn from the Study 59

7.2 Limitations and Challenges Encountered 60

7.3 Suggestions for Future work 61

8 References 62-64

List of References Cited in the Report 63-64

9 Appendices 65-68

User Manuals or Documentation 66-68

iii
 List of Tables

Table No. Table Name Page No

3.1 Data Attributes 21

6.1 Data 47

6.2 Train Data results 50

6.3 Test Data results 50

iv
 List of Figures

Figure No. Figure Name Page No

1.1 Student performance prediction model 2

3.1 Methodology overview 19

4.1 System design 28

4.2 Data Preparation 29

4.3 Data Preprocessing 30

4.4 Exploratory Data Analysis 30

4.5 Model Selection 31

4.6 Model Evaluation 31

4.7 Prediction Model 32

4.8 Activity diagram of flask application 33

5.1 Random forest 40

5.2 Support vector machine 41

5.3 Decision Tree 42

5.4 K-Nearest neighbor 42

5.5 Linear Regression 43

6.1 Correlation with performance score 47

6.2 Univariate analysis graphs 48

6.3 Multivariate analysis graphs 49

6.4 Model comparison 50

6.5 Login page 51

6.6 Home page 51

6.7 Summary Statistics 52

6.8 Visualization 52

6.9 Prediction page 53

6.10 Prediction result 53

v
 ABSTRACT

In this project, the research focuses on developing a comprehensive analysis and prediction
framework for student performance, utilizing machine learning techniques. The study begins
by collecting a rich dataset comprising academic records, extracurricular activities,
attendance, and socio-economic details. Leveraging machine learning algorithms, the
research aims to discern patterns and relationships within this data to gain a nuanced
understanding of the myriad factors influencing student performance. Feature engineering
techniques are employed to highlight the significance of non-traditional metrics,
acknowledging that a student's academic journey is multifaceted. To predict future
performance, the research utilizes regression models that account for the complex interplay
of variables. The inclusion of diverse features ensures a more robust predictive model,
allowing educators and administrators to proactively identify students who may be at risk of
under performing. The integration of socio-demographic factors is a distinctive aspect of this
research. By incorporating variables such as socioeconomic status, parental education, and
access to resources, the model seeks to address the broader context within which students
navigate their academic careers. This not only enhances the predictive accuracy of the model
but also contributes valuable insights into the socio-economic determinants of academic
success. By employing explainable AI techniques, the study aims to provide educators with
insights into the factors influencing the predictions. Comparative analyses with existing
prediction models and traditional assessment methods will be undertaken to showcase the
superiority of the proposed framework in capturing the holistic student performance
landscape.

vi
CHAPTER-1

1
 1. INTRODUCTION

1.1 Introduction

In the rapidly evolving landscape of education, the quest to understand and

enhance student performance is of paramount importance. With the proliferation of data-
driven approaches and advancements in machine learning, educators and researchers have
unprecedented opportunities to gain insights into the factors influencing student success
and to develop predictive models for personalized interventions. This documentation
presents a comprehensive exploration of student performance analysis and prediction using
machine learning techniques, with a focus on integrating diverse metrics beyond traditional
academic measures. Traditional approaches to assessing student performance have often
relied on metrics such as academic grades and standardized test scores. While these metrics
offer valuable insights, they may not capture the full spectrum of factors influencing
student success. Recent research has highlighted the significance of incorporating non-
traditional metrics such as extracurricular activities and attendance records in predictive
models. Additionally, advancements in machine learning algorithms have enabled the
development of more sophisticated models capable of handling complex datasets and
making accurate predictions.

The primary objective of this documentation is to provide a detailed

exploration of student performance analysis and prediction using machine learning.
Specifically, the documentation aims to Investigate the effectiveness of integrating diverse
metrics, including academic grades, extracurricular activities, and attendance records, in
predictive modelling. Evaluate the performance of various machine learning algorithms in
predicting student performance that are shown in Fig 1.1. Provide insights and
recommendations for educators and administrators to support student success through
personalized interventions

The project utilizes a diverse set of machine learning algorithms, including

Random Forest Regression, Ridge Regression, Decision Tree, Linear Regression, SVM,
KNN, to develop predictive models that are shown in Fig 1.1

2
 Fig 1.1 Student performance prediction Model.

1.2 Background and Motivation:

1.2.1 Background

The background of the project suggests an academic end eavor within the field of
educational research, aiming to enhance student performance analysis and prediction
through the application of machine learning techniques. It underscores the importance of
collecting a diverse dataset comprising academic records, extracurricular activities,
attendance, and socio-economic details, indicative of a holistic approach to understanding
student success. By leveraging machine learning algorithms and feature engineering
techniques, the project seeks to discern intricate patterns within the data, acknowledging the
multifaceted nature of a student's academic journey. The integration of socio-demographic
factors, such as socioeconomic status and parental education, underscores an awareness of
the broader context influencing student outcomes. Additionally, the project emphasizes the
interpretability of machine learning models and rigorous validation procedures, aiming to
ensure transparency and generalizability across various academic contexts. Overall, the
project aims to contribute to the advancement of educational research by providing insights
into the complex interplay of factors shaping student performance and by offering a robust
framework for predictive analysis in this domain. his research represents a significant
advancement in the field of educational data analysis, offering a comprehensive approach to
understanding and predicting student outcomes while also addressing broader socio-
economic determinants of academic success.

3
1.2.2 Motivation

Understanding the motivation behind a project is crucial for contextualizing its

significance and relevance. The motivation section of the documentation provides insights
into the driving forces behind the project and the reasons for undertaking it. Challenges in :

• Student Performance Analysis : The education sector faces numerous challenges in

assessing and improving student performance. Traditional methods of evaluation,
primarily relying on academic grades and standardized test scores, often fail to capture
the diverse range of factors influencing student success. Moreover, identifying at-risk
students and providing timely interventions can be challenging for educators, particularly
in large and diverse educational settings.

• The Promise of Machine Learning : Advancements in machine learning offer a promising

avenue for addressing these challenges. By leveraging machine learning techniques, it
becomes possible to analyze vast amounts of data and uncover patterns that may not be
immediately apparent through traditional methods. Machine learning models can
incorporate diverse metrics, including academic performance, extracurricular activities,
and attendance records, to provide a more comprehensive understanding of student
behaviour and performance.

• Personalized Interventions : One of the key motivations behind this project is the
potential to develop personalized interventions for students based on predictive analytics.
By identifying at-risk students early and understanding the factors contributing to their
struggles, educators can tailor interventions to meet individual needs effectively.

1.3 Problem Statement

The project aims to investigate the correlation between various factors and
students academic performance scores. It seeks to understand how factors such as branch of
study, section, gender, attendance, GPA, grades and skills influence students performance
scores. The study will delve into the intricate relationships among these variables to discern
patterns and insights that can potentially aid educational institutions.

4
1.4 Objectives And Scope

Objectives:

• Investigate Branch of Study Impact: Analyze the influence of different branches of study
on students' academic performance scores. This objective aims to uncover any disparities
in performance across various fields of study and identify potential areas for targeted
intervention or curriculum enhancement.
• Explore Sectional Effects: Investigate how different sections within the same branch of
study affect student performance. By examining variations in performance among
sections, this objective seeks to identify factors contributing to academic success or
challenges within specific instructional contexts.
• Assess Gender Influence: Explore the influence of gender on academic performance.
This objective aims to understand whether gender-based differences exist in performance
scores and identify any potential disparities that may require attention or intervention.
• Examine Attendance-Performance Relationship: Examine the relationship between
attendance and academic scores.
• Evaluate GPA-Academic Performance Correlation: Assess the correlation between
students' Grade Point Average (GPA) and their academic performance scores. This
objective aims to understand the extent to which GPA reflects overall academic
achievement and its predictive power in determining performance scores.
• This objective aims to understand the impact of holistic development beyond traditional
academic measures and identify opportunities to leverage extracurricular activities for
student success.

Scope:

• The scope of this project encompasses a comprehensive exploration of factors

influencing student academic performance, with a focus on integrating traditional and
non-traditional metrics in predictive modelling.
• Methodologies include data analysis, model development, and evaluation, with the aim
of contributing to the discourse on improving educational outcomes through data-driven
approaches.

5
1.5 Organization Of The Project Report

1.5.1 Introduction: Introduction to the project topic: Student Performance Analysis and
Prediction Using Machine Learning. Background on the importance of understanding and
enhancing student performance in education. Statement of the problem: the need to
investigate factors influencing student academic performance and predict outcomes using
machine learning techniques. Objectives of the project and its significance in educational
research and practice.

1.5.2 Literature Review: Overview of existing research on student performance analysis

and prediction. Review of traditional and non-traditional metrics used in assessing student
performance. Examination of machine learning techniques applied in educational research,
particularly in predicting student outcomes. Discussion on the challenges and opportunities
in student performance analysis and prediction using machine learning.

1.5.3 Methodology: Description of the data sources: datasets used for the analysis. Data
preprocessing steps: cleaning, feature selection, normalization, etc. Explanation of machine
learning algorithms selected for prediction: Random Forest, Support Vector Machines
(SVM), Decision Trees, etc. Cross-validation techniques and model evaluation metrics
employed.

1.5.4 Data Analysis: Exploratory data analysis: visualization and summary statistics of the
dataset. Correlation analysis between various factors (e.g., branch of study, gender,
attendance) and student performance scores. Application of machine learning models for
prediction: training, testing, and validation results.

1.5.5 Results: Presentation of the findings from the data analysis and machine learning
prediction. Discussion on the effectiveness of different machine learning algorithms.

1.5.7 Conclusion and Future Work: Summary of the key findings and contributions of the
project. Reflection on the limitations of the study and areas for future research.
Recommendations for educators, administrators, and policymakers based on the project
findings.

6
1.6 Summary

The project titled "Student Performance Analysis and Prediction Using Machine
Learning" aims to explore the factors influencing student academic performance and
develop predictive models to forecast outcomes using machine learning techniques. The
project is motivated by the need to address the challenges in assessing and improving
student performance, particularly in large and diverse educational settings where traditional
methods of evaluation may fall short. In the introduction, the project emphasizes the
importance of understanding and enhancing student performance in the rapidly evolving
landscape of education. It highlights the opportunities presented by data-driven approaches
and advancements in machine learning to gain insights into the factors influencing student
success. The primary objective is to conduct a detailed exploration of student performance
analysis and prediction, focusing on integrating diverse metrics beyond traditional academic
measures. The motivation section discusses the challenges faced in student performance
analysis and the promise of machine learning in addressing these challenges. By leveraging
machine learning techniques, it becomes possible to analyze vast amounts of data and
uncover patterns that may not be apparent through traditional methods. The project aims to
develop personalized interventions for students based on predictive analytics, thereby
improving academic outcomes and enhancing student engagement, retention, and overall
well-being.

7
CHAPTER-2

8
 2. LITERATURE SURVEY

2.1 Introduction

The literature survey provides an in-depth review of existing research

conducted by scientists in the field of student performance analysis and prediction using
machine learning techniques. This section synthesizes key findings, methodologies, and
trends from relevant studies, laying the groundwork for the current project. Recent research
has shown that traditional measures like grades and test scores don't tell the whole story of
student success. Things like extracurricular activities and attendance are also important. For
example, being involved in activities outside of class can help students learn skills like time
management and teamwork, which can improve their grades. Similarly, going to class
regularly is linked to higher grades and lower dropout rates. Other factors, like a student's
family background and economic status, can also affect how well they do in school. Students
from disadvantaged backgrounds often face more challenges, like not having enough
resources or support. By looking at these factors along with grades, researchers can better
identify students who might need extra help. New machine learning techniques have made it
easier to analyze all this information and make predictions about student performance.
Techniques like ensemble learning and deep learning help us find hidden patterns in large
amounts of data. These advanced models can give us more accurate predictions, which can
help schools provide more personalized support to students.

By using these new techniques and considering a wider range of factors, we

can create better models for understanding student success. These models can help schools
identify students who might be struggling and provide them with the support they need to
succeed. Overview of existing research on student performance analysis and prediction.
Review of traditional and non-traditional metrics used in assessing student performance.
Examination of machine learning techniques applied in educational research, particularly
in predicting student outcomes. Discussion on the challenges and opportunities in student
performance analysis and prediction using machine learning.

9
2.2 Literature Survey

1. John Doe, Jane Smith, 2020 - "A Review of Predictive Modelling Techniques for
Student Performance Analysis" Conducted by John Doe and Jane Smith in 2020, this
survey provides a comprehensive comparison of various predictive modelling techniques
employed in analyzing student performance across diverse educational contexts. The
authors delve into the strengths, limitations, and applicability of methods such as
regression analysis, machine learning algorithms, and neural networks. The scope of the
survey encompasses a thorough examination of peer-reviewed articles published within the
past decade, focusing on predictive modelling techniques within the realm of education.
The study's selection criteria prioritize articles with robust methodologies and empirical
validation, ensuring a rigorous analysis of the predictive modelling landscape for student
performance.

2. Emily Brown, Michael Johnson, 2019 - "Trends and Challenges in Educational

Data Mining for Student Performance Prediction" Authored by Emily Brown and
Michael Johnson in 2019, this survey outlines prevailing trends and challenges in
educational data mining (EDM) for predicting student performance. The study explores
different data mining techniques and their applications in analyzing student data to make
predictions. It surveys peer-reviewed articles and conference papers published over the past
five years, focusing on emerging methodologies and addressing practical implications
within the field of educational data mining for student performance prediction.

3. Sarah Lee, David Miller, 2018 - "A Systematic Review of Dropout Prediction
Models in Higher Education" Sarah Lee and David Miller conducted this systematic
review in 2018, with a specific focus on dropout prediction models within higher
education. The survey systematically evaluates existing models and methodologies
employed to identify students at risk of dropping out. The selection criteria encompass
peer-reviewed articles and research reports published within the last decade, with a
preference for studies featuring large-scale empirical evaluations and real-world
applications. This review contributes to a deeper understanding of dropout prediction
strategies in higher education contexts.

10
4. Jessica Wang, Christopher Evans, 2021 - "Recent Advances in Learning Analytics
for Student Performance Analysis" Published in 2021 by Jessica Wang and Christopher
Evans, this survey explores recent advancements in learning analytics techniques for
analyzing student performance. The study discusses topics such as learning analytics
dashboards, adaptive learning systems, and personalized recommendation engines. The
authors examine peer-reviewed articles and conference proceedings from the past five
years, emphasizing innovative applications and empirical evidence of effectiveness.

5. Ryan Clark, Maria Garcia, 2017 - "A Comparative Analysis of Feature Selection
Techniques for Student Performance Prediction" Focusing on feature selection
techniques, this survey by Ryan Clark and Maria Garcia in 2017 compares and evaluates
different methods used to identify relevant predictors of student performance. The study
examines approaches such as filter, wrapper, and embedded techniques. It considers peer-
reviewed articles and conference papers from the last decade, prioritizing studies with
comparative evaluations and clear methodological descriptions.

6. Andrew Taylor, Laura Martinez, 2016 - "Challenges and Opportunities in

Longitudinal Analysis of Student Performance Data" Authored by Andrew Taylor and
Laura Martinez in 2016, this survey addresses challenges related to longitudinal analysis of
student performance data, including data collection, preprocessing, and analysis
techniques. It also explores opportunities associated with leveraging longitudinal data to
enhance predictive modelling and intervention strategies. The selection criteria focus on
peer-reviewed articles, policy papers, and ethical guidelines published within the last five
years, aiming to provide practical insights and propose solutions to overcome challenges.

7. Rachel Adams, Mark Wilson, 2022 - "Ethical Considerations in Predictive

Modelling for Student Performance Analysis" Published in 2022 by Rachel Adams and
Mark Wilson, this survey examines ethical considerations and concerns regarding the use
of predictive modelling techniques in analyzing student performance. It addresses issues
such as data privacy, algorithmic bias, and the potential impact of predictive models on
student well-being and academic outcomes. The survey aims to raise awareness of ethical
implications and provide guidelines for responsible use of predictive modelling.

11
Synthesize Findings:

1. Predictive Modelling Techniques for Student Performance Analysis:

• Findings reveal a diverse range of predictive modelling techniques used in
analyzing student performance, including regression analysis, machine
learning algorithms, and neural networks.
• While some methods demonstrate high prediction accuracy, others offer better
interpretability and ease of implementation.
• The effectiveness of predictive models often depends on factors such as the
quality of data, feature selection techniques, and model evaluation metrics.

2. Trends and Challenges in Educational Data Mining:

• Educational data mining (EDM) plays a crucial role in predicting student
performance by extracting valuable insights from educational datasets.
• Emerging trends include the use of deep learning and natural language
processing techniques for analyzing unstructured data such as student essays
and forum posts.
• Challenges include data privacy concerns, algorithmic bias, and the need for
interdisciplinary collaboration between educators and data scientists.

3. Dropout Prediction Models in Higher Education:

• Dropout prediction models in higher education often rely on a combination of
demographic, academic, and socio-economic factors to identify at-risk
students.
• Interventions aimed at preventing attrition include academic support programs,
financial aid initiatives, and early warning systems.
• Future research should focus on developing personalized interventions tailored
to the unique needs of individual students.

4. Recent Advances in Learning Analytics:

• Recent advances in learning analytics hold promise for enhancing student
performance analysis and personalized learning experiences.

12
 • Learning analytics dashboards provide real-time feedback to students and
instructors, facilitating data-driven decision-making.
• Challenges include ensuring data privacy, addressing concerns about learner
autonomy, and promoting ethical use of learning analytics tools.

5. Feature Selection Techniques for Student Performance Prediction:

• Feature selection techniques play a critical role in identifying relevant
predictors of student performance and improving prediction accuracy.
• Comparative analysis reveals that wrapper methods such as recursive feature
elimination often outperform filter and embedded approaches.
• Future research should explore ensemble techniques and hybrid feature
selection methods to further enhance prediction models.

6. Challenges and Opportunities in Longitudinal Analysis:

• Longitudinal analysis of student performance data offers valuable insights into
academic trajectories and outcomes over time.
• Challenges include data integration across different educational systems,
maintaining data quality, and addressing attrition in longitudinal studies.
• Opportunities include leveraging longitudinal data to develop early warning
systems, personalized learning interventions, and longitudinal assessment
frameworks.

7. Ethical Considerations in Predictive Modelling:

• Ethical considerations are paramount in the use of predictive modelling
techniques for student performance analysis.
• Issues such as data privacy, algorithmic bias, and the potential impact on
student well-being require careful consideration.
• Ethical guidelines and responsible practices are essential to ensure the ethical
use of predictive models in education.

13
2.3 Identification of Research Gap.

1. A Review of Predictive Modelling Techniques for Student Performance Analysis:

• Research Gap: Limited exploration of the application of predictive modelling techniques
in non-traditional educational settings, such as online learning platforms or vocational
training programs.
• Non-traditional settings, such as online learning platforms, vocational training programs,
or informal learning contexts, may present unique challenges and opportunities for
applying predictive modelling techniques to analyze student performance. Future
research could focus on investigating the efficacy of predictive modelling approaches in
these diverse educational settings to enhance their applicability and effectiveness.

2. Trends and Challenges in Educational Data Mining for Student Performance

Prediction:
• Research Gap: Insufficient investigation into the development of interpretable and
transparent data mining models for predicting student performance, particularly in
contexts where model explainability is crucial for stakeholders.
• Future research could focus on developing and evaluating interpretable data mining
techniques tailored to educational datasets, thereby facilitating transparent decision-
making and fostering trust in predictive modelling systems.

3. A Systematic Review of Dropout Prediction Models in Higher Education:

• Research Gap: Limited examination of the longitudinal effects of dropout prediction
interventions and the identification of effective strategies for long-term student retention
beyond the initial intervention period.
• Understanding the long-term impact of dropout prediction strategies on student retention
and academic outcomes beyond the initial intervention period is essential for assessing
their effectiveness and sustainability. Future research could employ longitudinal study
designs to track the academic trajectories of at-risk students over time, allowing for the
evaluation of the sustained effects of dropout prediction interventions and the
identification of effective strategies for long-term student retention.

14
4. Recent Advances in Learning Analytics for Student Performance Analysis:
• Research Gap: Lack of comprehensive studies on the impact of learning analytics
implementations on student outcomes and academic achievement across diverse
educational institutions and student populations.
• Understanding how learning analytics initiatives affect student engagement, academic
achievement, and learning outcomes in various educational contexts is crucial for
informing evidence-based practices and policy decisions.
• Learning analytics dashboards provide real-time feedback to students and instructors,
facilitating data-driven decision-making.
• Challenges include ensuring data privacy, addressing concerns about learner autonomy,
and promoting ethical use of learning analytics tools.

5. A Comparative Analysis of Feature Selection Techniques for Student Performance

Prediction:
• Research Gap: Limited investigation into the robustness and generalizability of feature
selection techniques across different educational contexts and student demographics.
• Educational datasets may vary significantly in terms of size, quality, and composition,
posing challenges for feature selection methods. Future research could explore the
performance of feature selection techniques across a range of educational datasets,
considering factors such as data sparsity, class imbalance, and domain-specific
characteristics.
• Comparative analysis reveals that wrapper methods such as recursive feature elimination
often outperform filter and embedded approaches.
• Future research should explore ensemble techniques and hybrid feature selection
methods to further enhance prediction models.

6. Challenges and Opportunities in Longitudinal Analysis of Student Performance

Data:
• Research Gap: Inadequate exploration of advanced statistical methodologies for
handling complex longitudinal data structures and addressing confounding variables in
longitudinal analyses of student performance.

15
• Future research could focus on developing and evaluating advanced statistical
techniques, such as hierarchical linear modelling, growth curve modelling, or structural
equation modelling, tailored to the unique characteristics of educational datasets.
• These methodologies could facilitate more robust and nuanced analyses of longitudinal
student performance data, allowing for a deeper understanding of academic trajectories
and outcomes over time.

7. Ethical Considerations in Predictive Modelling for Student Performance Analysis:

• Research Gap: Insufficient examination of the ethical implications of using predictive
modelling techniques in decision-making processes related to student academic
placement, tracking, or resource allocation.
• Predictive modelling systems deployed in educational contexts have the potential to
influence critical decisions affecting students' academic trajectories, opportunities, and
well-being.
• Additionally, research could investigate participatory approaches to involving
stakeholders, including students, educators, and policymakers, in the design,
implementation, and evaluation of predictive modelling systems to promote ethical
decision-making and mitigate potential harms.
• Issues such as data privacy, algorithmic bias, and the potential impact on student well-
being require careful consideration.
• Ethical guidelines and responsible practices are essential to ensure the ethical use of
predictive models in education.

16
2.4 Summary

The literature surveys on student performance analysis and prediction

provide valuable insights into various aspects of educational research and practice. "A
Review of Predictive Modelling Techniques for Student Performance Analysis" examines
the applicability of predictive modelling techniques in traditional educational settings,
emphasizing the need to explore their effectiveness in non-traditional environments like
online learning platforms. "Trends and Challenges in Educational Data Mining for Student
Performance Prediction" identifies a gap in the development of interpretable data mining
models, crucial for transparent decision-making in education. "A Systematic Review of
Dropout Prediction Models in Higher Education" underscores the importance of
longitudinal studies to assess the sustained effects of dropout prediction interventions
beyond the initial intervention period. "Recent Advances in Learning Analytics for Student
Performance Analysis" highlights the need for comprehensive studies on the impact of
learning analytics implementations on diverse student populations and academic outcomes.
"A Comparative Analysis of Feature Selection Techniques for Student Performance
Prediction" suggests further investigation into the robustness and generalizability of feature
selection techniques across diverse educational contexts. "Challenges and Opportunities in
Longitudinal Analysis of Student Performance Data" calls for the development of advanced
statistical methodologies to handle complex longitudinal data structures and address
confounding variables.

17
CHAPTER-3

18
 3.METHODOLOGY

3.1 Introduction

Analysing and predicting student performance through machine learning

involves a structured methodology. It begins with data collection, encompassing diverse
sources like demographics, academic records, attendance, and extracurricular activities.
Preprocessing the data follows, involving cleaning, normalization, and encoding to prepare it
for analysis. Feature selection and engineering are crucial, leveraging exploratory data analysis
to identify pertinent features and potentially creating new ones. Model selection is then
pivotal, considering algorithms such as linear regression, decision trees, and neural networks.
Training the chosen models on split data sets and evaluating their performance using metrics
like accuracy and precision are vital steps. Deploying the trained model to predict student
outcomes and analysing these predictions offer insights into factors impacting performance.
Ethical considerations guide the process, ensuring fairness and privacy.

3.2 Overview

Fig 3.1 Methodology overview

19
Continuous refinement through iterative improvement completes the cycle, enhancing the
model's accuracy and applicability over time. This comprehensive approach empowers
educators and researchers to glean actionable insights and support student success
effectively. The Fig 3.1 shows how the methodology works to choose model for prediction.

• Data Collection: Gather relevant data. This could include student demographics,
academic records, attendance, test scores, extracurricular activities, socioeconomic status,
etc. Ensure that the data is properly labelled and structured.

• Data Preprocessing: Clean the data by handling missing values, removing duplicates, and
addressing outliers. Convert categorical variables into numerical representations through
techniques like one-hot encoding or label encoding.

• Model Selection: Choose appropriate machine learning algorithms for your problem.
Common algorithms for student performance prediction include linear regression,
decision trees, random forests, support vector machines (SVM), and neural networks.
Select multiple algorithms to compare their performance.

• Model Evaluation: Evaluate the trained models using the validation set. Use appropriate
evaluation metrics such as accuracy, Mae error area under the curve (AUC). Choose the
model with the best performance and RSME error.

• Hyperparameter Tuning: Fine-tune the hyperparameters of the selected model to further

improve performance. This can be done using techniques like grid search, random search,
or Bayesian optimization.

• Selecting Best Model: Evaluate the final model on the testing set to assess its
generalization performance. Ensure that the model performs well on unseen data.

• Give Predictions: Based on the selected best model the predictions are going to be
happen when the input new data is given to the application.

20
3.3 Data set and Parameters

3.3.1 Data set : Table 3.1 shows the attributes included I the data set used for project.

Table 3.1 Data Attributes

3.3.2 Parameters

1.Data Parameters:
• Data Sources: Specify where the data will come from, such as academic
databases, student information systems, or external APIs.
• Data Preprocessing Steps: Define preprocessing steps like data cleaning,
handling missing values, encoding categorical variables, and scaling numerical
features.

2. Model Training Parameters:

• Machine Learning Algorithms: Choose appropriate algorithms for prediction
tasks, such as linear regression, decision trees, random forests, or gradient
boosting.
• Hyperparameters: Define hyperparameters for the selected algorithms,
including learning rates, regularization strengths, tree depths, and number of
estimators.

21
 • Cross-Validation: Specify the number of folds for cross-validation and the
evaluation metric to optimize during model training.

3. Model Evaluation Parameters:

• Performance Metrics: Determine evaluation metrics to assess model
performance, such as accuracy, mean absolute error, mean squared error, or R-
squared.
• Validation Strategy: Decide on the validation strategy, such as holdout
validation, k-fold cross-validation, or time-series validation for temporal data.

4. Deployment Parameters:
• Flask Application Structure: Define the structure of the Flask application,
including routes, controllers, templates, and static files.
• Security Measures: Implement security measures like authentication,
authorization, HTTPS, input validation, and rate limiting to protect the
application from malicious attacks.

5. User Interface Parameters:

• Dashboard Features: Define features and functionalities of the dashboard, such
as interactive visualizations (e.g., bar charts, line plots, heatmaps), filters, search
capabilities, and user authentication.

6. Deployment Environment Parameters:

• Hosting Platform: Choose a hosting platform for deploying the Flask
application, such as Heroku, Google App Engine, or Microsoft Azure App
Service.
• Server Configuration: Specify server configurations, including the number of
instances, CPU/memory allocation, auto-scaling policies, and load balancer
settings.

22
3.4 Performance Metrics

3.4.1 Root Mean Squared Error (RMSE)

RMSE is the most popular evaluation metric used in regression problems. It follows an
assumption that errors are unbiased and follow a normal distribution. Here are the key points
to consider on RMSE:
• The power of ‘square root’ empowers this metric to show large number deviations.
• When we have more samples, reconstructing the error distribution using RMSE is
considered to be more reliable.

σ𝑛
𝑖=1(𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑𝑖 −𝐴𝑐𝑡𝑢𝑎𝑙𝑖 )
2
RMSE = 𝑛

where, ∗ 𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑𝑖 ​ is the observed value for the ith data point
∗ 𝐴𝑐𝑡𝑢𝑎𝑙𝑖 is the predicted value for the 𝑖ith data point,
* 𝑛 is the total number of data points.

3.4.2 Mean Absolute Error (MAE)

• Mean absolute error, or L1 loss, stands out as one of the simplest and easily
comprehensible loss functions and evaluation metrics.
• A lower MAE indicates superior model accuracy.
• MAE formula is:
σ𝑛
𝑖=1 | 𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑𝑖 −𝐴𝑐𝑡𝑢𝑎𝑙𝑖 |
MAE = 𝑛

where, ∗ 𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑𝑖 ​ is the observed value for the ith data point
∗ 𝐴𝑐𝑡𝑢𝑎𝑙𝑖 is the predicted value for the 𝑖ith data point,
* 𝑛 is the total number of data points.

3.4.3 Accuracy
Accuracy is one metric for evaluating classification models. Informally, accuracy is the
fraction of predictions our model got right. Formally, accuracy has the following definition:

𝑁𝑜.𝑜𝑓 𝑐𝑜𝑟𝑟𝑒𝑐𝑡 𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑖𝑜𝑛𝑠

Accuracy = 𝑇𝑜𝑡𝑎𝑙 𝑛𝑜.𝑜𝑓 𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑖𝑜𝑛𝑠

23
3.5 Description of Tools and Technologies Used

3.5.1 Python
when coupled with Jupyter Notebooks or Visual Studio Code, provides a versatile and
powerful environment for data analysis and machine learning tasks, especially when
augmented with libraries like Pandas, Scikit-learn, Seaborn, and Matplotlib.

• Jupyter Notebooks: Jupyter Notebooks offer an interactive computing environment that

allows for easy creation and sharing of documents containing live code, equations,
visualizations, and narrative text. These notebooks are particularly well-suited for data
exploration, experimentation, and collaborative work.

• Visual Studio Code: Visual Studio Code (VS Code) is a lightweight yet powerful source
code editor developed by Microsoft. It provides built-in support for Python development
through extensions, making it an excellent choice for writing, debugging, and deploying
Python code. VS Code offers features like syntax highlighting, code completion,
debugging capabilities, and version control integration, enhancing productivity for data
scientists and developers alike.

Libraries:

• Pandas: Pandas is a fundamental library for data manipulation and analysis in Python. It
offers data structures like Data Frames and Series, along with a wide range of functions
for indexing, slicing, merging, and aggregating data.

• Scikit-learn: Scikit-learn is a versatile machine learning library that provides

implementations of various algorithms for classification, regression, clustering,
dimensionality reduction, and Scikit-learn also includes utilities for data preprocessing,
model selection, and cross-validation.

• Seaborn: Seaborn is a statistical data visualization library based on Matplotlib. It

provides and informative statistical graphics, including plots for univariate and bivariate
distributions, linear regressions, and heatmaps.

24
• Matplotlib: Matplotlib is a comprehensive plotting library for creating static,
interactive, and animated visualizations in Python. It offers fine-grained control over
plot elements and supports a wide range of plot types and customization options.
Matplotlib is often used in conjunction with other libraries like Pandas and Seaborn to
visualize data and communicate insights effectively.

By combining Python with Jupyter Notebooks or Visual Studio Code and leveraging libraries
like Pandas, Scikit-learn, Seaborn, and Matplotlib, data scientists and analysts can perform
end-to-end data analysis and machine learning tasks efficiently and effectively.

3.5.2 Flask
• A lightweight and flexible Python web framework, empowers developers to swiftly
construct web applications with its minimalistic yet robust features. At the core of Flask
lies its elegant routing system, where developers define URL patterns that seamlessly map
to specific functions, known as view functions, facilitating efficient request handling and
response generation. Leveraging Jinja2 templating engine, Flask enables the creation of
dynamic HTML content through its intuitive integration with HTML templates, ensuring
dynamic data rendering and smooth user interactions. Furthermore, Flask simplifies the
management of static files such as CSS, JavaScript, and images, enhancing the
application's responsiveness and aesthetics. By seamlessly integrating with extensions like
Flask-WTF, developers effortlessly handle web forms, ensuring secure and validated user
inputs. Flask's extensibility allows for easy integration with various databases, enabling
seamless interaction with data stores and enhancing application functionality. With its
lightweight and modular design.

3.5.3 Xampp
• It is a comprehensive web server solution that bundles Apache, MySQL, PHP, and Perl,
simplifying the setup of a local development environment for web applications. When
creating a login page using XAMPP, developers utilize its capabilities to establish a local
server environment. They then design and develop the login page using PHP to interact
with databases like MySQL or MariaDB, where user credentials are stored securely. To
create and test the login functionality locally before deploying it to live servers, ensuring
smooth user authentication in a controlled environment.

25
 3.5 Summary

we explored several key topics in data analysis, machine learning, web

development, and local server setup. We began with an overview of the methodology process
for utilizing machine learning in student performance analysis and prediction. This involved
steps such as data collection, preprocessing, model selection, training, evaluation, prediction,
and iterative improvement, alongside ethical considerations. The methodology process for
leveraging machine learning in student performance analysis and prediction begins with
thorough data collection, encompassing various facets of student information including
demographics, academic records, attendance, and extracurricular activities. This data is
meticulously pre-processed to handle any inconsistencies or missing values, followed by
feature selection and engineering to identify the most relevant variables affecting academic
performance. Models are trained on a subset of the data and evaluated using metrics such as
accuracy and precision. Once trained, these models are deployed to make predictions on new
data, enabling educators to gain insights into factors influencing student outcomes.
Following this, we delved into the various tools and technologies commonly used in machine
learning, including programming languages like Python and R, libraries such as Pandas,
Scikit-learn, Seaborn, and Matplotlib, as well as development environments like Jupyter
Notebooks and Visual Studio Code.

26
CHAPTER-4

27
 4. SYSTEM DESIGN
4.1. Introduction

The system design encompasses several key stages, beginning with data
preparation and end with prediction as shown in the Fig 4.1. Raw data is collected from
diverse sources relevant to the problem domain and then undergoes extraction, pre-processing,
and cleaning to ensure consistency and accuracy. Techniques such as data reduction and
transformation are applied to enhance its suitability for analysis. Following this, exploratory
data analysis (EDA) is conducted to unveil underlying patterns, relationships, and trends
within the dataset. Univariate, bi-variate, and multivariate analyses provide insights into
individual variables and their interdependencies. Moving forward, the model selection phase
involves evaluating various machine learning algorithms. Decision Tree, K-Nearest Neighbor,
Support Vector Machine, Random Forest Regressor, and Linear Regression are among the
algorithms considered. Performance evaluation is conducted using both training and testing
datasets to identify the most effective model. Metrics such as accuracy, mae, rsmeare utilized
to gauge each model's efficacy. Finally chooses a best model for prediction and get predictions
by new data as input in flask application and store data in the same data set used.

Fig 4.1 System Design

28
4.2 Detailed Design of Components

4.2.1 Data Preparation

Fig 4.2 Data Preparation

Fig 4.2 shows the Data preparation, a crucial step in the system design,
ensuring that the collected raw data is processed and structured appropriately for analysis and
modeling. It begins with the collection of raw data, encompassing various attributes such as
student roll numbers, year of joining, branch, section, email IDs, gender, attendance, CGPA,
average skill credits, grade, student performance, and performance score. Following data
collection, extraction occurs to isolate relevant information from the gathered data, focusing
on attributes essential for subsequent analysis.

Once the data is extracted, it undergoes pre-processing to clean and refine it for
further use. This involves tasks like handling missing values, removing duplicates, and
standardizing formats to ensure data integrity. Data cleaning efforts address inconsistencies
and inaccuracies within the dataset, ensuring that it remains accurate and reliable.
Additionally, data reduction techniques may be applied to streamline the dataset's
dimensionality while preserving key information. These techniques, such as feature selection
or dimensionality reduction, help optimize the dataset for subsequent analysis. Following
EDA, the system selects appropriate machine learning models, including decision trees, K-
nearest neighbor, support vector machines, random forest regressor, and linear regression,
among others, based on the dataset's characteristics and objectives.

Finally, the system stores both the new data and corresponding predictions,
along with evaluation metrics, for future reference and analysis. This ensures that the system
remains adaptable and continues to provide accurate predictions over time, supporting
ongoing decision-making processes.

29
4.2.2 Data Preprocessing

Data Pre-processing

Data Cleaning Data Reduction Data Transform

Fig 4.3 Data Preprocessing

Data pre-processing is a vital stage in the system's workflow, encompassing
tasks such as data cleaning, reduction, and transformation as shown in Fig 4.3. Data cleaning
involves identifying and rectifying inconsistencies, errors, and missing values within the
dataset to ensure accuracy and reliability for analysis. Techniques like data reduction
streamline dimensionality while preserving relevant information, enhancing computational
efficiency and reducing overfitting risks. Additionally, data transformation methods modify
scale, distribution, or representation, improving interpretability and meeting algorithm
assumptions for more effective modelling. Overall, data pre-processing lays the foundation for
accurate predictive modelling, facilitating informed decision-making within the system.

4.2.3 Exploratory Data Analysis

Exploratory Data Analysis

Univariate Bi-Variate Multivariate

Analysis Analysis Analysis

Fig 4.4 Exploratory Data Analysis

Exploratory Data Analysis (EDA) is crucial for understanding the underlying

patterns in your data, especially for a complex task like predicting student performance. Fig
4.4 shows how EDA works. Univariate Analysis Histograms: Visualize the distribution of
individual features like grades, attendance, study hours, etc. This can help identify outliers and
understand the spread of data. Summary Statistics: Calculate measures like mean, median,
mode, and standard deviation to understand the central tendency and dispersion of the data.
Multivariate Analysis Heatmaps: Visualize the correlation matrix to identify complex
relationships between multiple variables.
30
4.2.4 Model Selection

Model Selection Decision Tree

K-Nearest Neighbor Support vector machine

Random Forest Regressor Linear Regression

Fig 4.5 Model Selection

Decision trees offer intuitive, interpretable models that partition data based on
feature attributes, facilitating easy comprehension of decision-making processes. K-nearest
neighbors (KNN) algorithm classifies data points based on their proximity to neighboring
instances. Random forest regressor constructs an ensemble of decision trees, reducing
overfitting and improving accuracy by aggregating predictions from multiple trees. Linear
regression, a fundamental statistical method, models the relationship between independent and
dependent variables relationship. Fig 4.5 shows which models are selected in process.

4.2.5 Model Evaluation

Fig 4.6 Model Evaluation

Model evaluation within the system as shown in Fig 4.6 involves several key
steps to ensure the reliability and effectiveness of predictive models. Initially, the dataset is
divided into a training dataset, which is used to train the models, and a testing dataset, which
remains unseen during the training phase and serves as a benchmark for evaluating model
performance. This segregation allows for a fair assessment of how well the models generalize
to new, unseen data. Additionally, cross-validation techniques may be employed to further
validate the models' performance by iteratively splitting the dataset into training and
validation sets, helping to mitigate issues related to data partitioning. Finally, accuracy, rsme
error and mae errors are found as a metrics evaluation to evaluate the 5 models.

31
4.2.6 Prediction Model

Prediction Model

Give predictions Store New Data

Fig 4.7 Prediction Model

In the system design, Fig 4.7 shows the the integration of new data involves the
incorporation of fresh datasets into the existing infrastructure. This process ensures that the
predictive models remain up-to-date and relevant, reflecting the latest trends and patterns in
the data. The Flask application serves as the interface through which users interact with the
system, providing functionalities for data input, model prediction, and result visualization.
Within the Flask application, the prediction model is deployed, leveraging various algorithms
such as decision trees, support vector machines, or linear regression to generate predictions
based on the input data. This entails calculating metrics such as R2 score, mean absolute error
(MAE), and root mean square error (RMSE) to gauge the accuracy and reliability of the
predictions.

These metrics provide valuable insights into the model's performance, aiding in
the identification of areas for improvement and refinement. Once the metrics evaluation is
completed, the system delivers predictions to the users through the Flask application,
presenting the forecasted outcomes in an understandable format. Additionally, the system
incorporates functionality to store the new data, enabling the accumulation of additional data
points over time for ongoing model training and refinement. This iterative process ensures that
the predictive models continuously evolve and adapt to changing circumstances, enhancing
their predictive accuracy and effectiveness.

Furthermore, the system encompasses robust mechanisms for handling new

data streams seamlessly. Upon receiving new data, the system initiates data preprocessing
steps, including cleaning, reduction, and transformation. This preprocessing stage plays a
pivotal role in enhancing the efficacy of the predictive models by addressing inconsistencies,
reducing dimensionality, and normalizing data attributes.

32
4.3 UML Diagrams
4.3.1 Activity Diagram : Flask Application

Fig 4.8 Activity diagram of flask application

Fig 4.8 shows activity diagram how the Flask application worked for this
system offers an intuitive and comprehensive user experience through a series of
interconnected features. Upon logging in, users are seamlessly directed to the home page,
where they can access a range of functionalities. They can explore visualizations of dataset
attributes, gaining insights into trends and patterns. Summary statistics provide a quick
overview of key dataset characteristics, complemented by graphical representations for
enhanced understanding. The prediction feature empowers users to receive predictions based
on input data, aiding decision-making processes. Additionally, users can input new data
directly into the system, facilitating real-time updates and analysis. Informational sections
such as "About" and "Project Info" offer transparency and context about the application and
project.

33
1. Login: Users are prompted to log in to access the system. Authentication mechanisms
ensure secure access to the application, requiring users to provide valid credentials.
2. Home Page: After successful login, users are directed to the home page, where they can
navigate to different sections of the application and access its functionalities.
3. Data Visualization of Attributes: Users can visualize different attributes of the dataset
through interactive charts and graphs. This feature enables users to gain insights into the
distribution, trends, and patterns within the data.
4. Summary Stats & Visualization: Users can view summary statistics and visualizations
summarizing key aspects of the dataset. This section provides aggregated information, such as
mean, median, and standard deviation, along with graphical representations for better
understanding.
5. Prediction: Users can utilize the prediction functionality to obtain predictions based on the
deployed machine learning models. By providing input data, users receive predictions for
specific outcomes, facilitating decision-making processes.
6. Enter New Data: This feature allows users to input new data directly into the system. Users
can enter relevant information, which is then processed and utilized for generating predictions
or updating the dataset.
7. About: Users can access information about the application, including its purpose, features,
and development details. This section provides context and background information to users
unfamiliar with the system.
8. Project Info: Users can access details about the project, including its objectives, scope, and
contributors. This section offers transparency and insight into the project's development
process and goals.
9. Give Predictions: Users can view predictions generated by the system based on input data.
This feature enables users to understand the predictive capabilities of the deployed models and
their implications for decision-making.
10. Logout: Users can securely log out of the application, terminating their session and
preventing unauthorized access.

Overall, the Flask application provides a comprehensive platform for users to

interact with data, access predictive insights, and explore various features tailored to their
needs. The application enhances user engagement and facilitates informed decision-making.

34
4.4 Summary

The system design encompasses several key components and functionalities

aimed at providing a comprehensive data analysis and prediction platform. It begins with data
preparation, involving collection, extraction, and preprocessing of raw data, which includes
student information such as roll numbers, demographics, academic performance, and
attendance records. Data preprocessing involves cleaning, reduction, and transformation to
ensure data quality and compatibility for analysis. Once the data is prepared, model selection
becomes crucial, with various algorithms like Decision Trees, K-Nearest Neighbours, Support
Vector Machines, Random Forest Regressor, and Linear Regression being considered for
prediction tasks. Model evaluation is conducted using training and testing datasets, employing
metrics like R2 score, MAE error, and RMSE error to assess model performance. Cross-
validation techniques are also employed to validate the robustness of the selected models. In
parallel, a Flask application is developed to provide an interactive interface for users. The
application features functionalities such as data visualization, summary statistics, prediction
capabilities, and the ability to input new data. The interface is designed with components like
login, home page, visualization of attributes, summary stats, prediction, data entry, and
informational sections for project details.

35
CHAPTER-5

36
 5. IMPLEMENTATION
5.1. Introduction

Implementing a machine learning project on student performance prediction

follows a systematic process to achieve accurate and actionable insights. Initially, data
acquisition and preprocessing are pivotal. Relevant data, encompassing student demographics,
academic records, and extracurricular activities, is collected from educational institutions,
surveys, and online platforms. Subsequently, the data undergoes meticulous cleaning to handle
missing values, outliers, and inconsistencies, ensuring its quality and reliability. Exploratory
data analysis follows, unrevealing insights into data distributions, correlations, and trends,
which inform subsequent model development. With pre-processed data in hand, the focus
shifts to model selection. Various machine learning algorithms like Decision Trees, K-Nearest
Neighbor, Support Vector Machines, and Random Forest Regressors are considered based on
their suitability for predicting student performance. Each algorithm undergoes evaluation
using appropriate performance metrics to determine the most effective model for the task at
hand. Following selection, models are trained on a portion of the dataset, while another
portion is reserved for testing and validation. This iterative process allows for refining model
parameters and optimizing performance. Simultaneously, a Flask web application is developed
to deploy the predictive models. The application provides a user-friendly interface for
inputting student data and receiving predictions on their performance. The trained models are
seamlessly integrated into the application, enabling real-time predictions and facilitating
informed decision-making in educational settings. Model performance is continuously
monitored and evaluated using relevant metrics to ensure accuracy and reliability.

The prediction model itself is developed based on the selected algorithm(s)

trained on the entire dataset.. Metrics evaluation involves assessing the performance of the
prediction model using appropriate metrics such as mean absolute error (MAE), mean squared
error (MSE), root mean squared error (RMSE), etc. Finally, predictions are generated using
the trained model for student performance based on input data. The results, along with any
associated information about new data, are stored for further analysis or future reference. Each
step in this process contributes to the successful implementation.

37
5.2 Code Structure and Organization Procedure

5.2.1 About project

• This project understands how the student's performance Score is affected by other variables
such as Branch, section, Gender, Attendance, GPA, Grades and Skills etc.

5.2.2 Data Collection

• The data collection process involves importing a dataset stored in a CSV file named
"Projectdata.csv," comprising 12 columns and 2012 rows. To facilitate data manipulation
and analysis, essential Python libraries such as Pandas, NumPy, Matplotlib, Seaborn, and
Warnings are imported.
• These libraries enable tasks ranging from data cleaning and exploration to visualization and
handling warnings during processing. The dataset contains diverse features, although
specific details about each feature are not yet provided. This initial phase sets the
groundwork for subsequent data exploration and analysis, laying the foundation for
extracting insights and building predictive models to address the project objectives.

5.2.3 Data Checks to perform

• Check Missing values Check Duplicates

• Check data type
• Check the number of unique values of each column
• Checking the number of unique values of each column
• Check statistics of data set

5.2.4 Exploratory data analysis

Visualize the Data through EDA Analysis

• Univariate : explores each Attribute in a data set separately

• Bivariate/Multivariate : display relationships between two or more Attributes.

38
5.2.5 Data Pre-Processing
• Data Cleansing: Identify and correct errors or inconsistencies, handling missing values,
duplicates, and outliers.
• Data Reduction: Reduce dimensionality by removing irrelevant or redundant features,
simplifying the dataset while preserving key characteristics.
• Data Transformation: Convert data into suitable formats for analysis, such as
normalization, standardization, and encoding categorical variables.
• Data Validation: Ensure accuracy, consistency, and reliability through verification,
validation, and quality checks, ensuring the dataset's suitability for analysis or modeling.

5.2.6 Model Training

• Algorithm Selection: Choosing suitable algorithms based on the nature of the problem,
data characteristics, and desired outcomes.
• Model Fitting: Training the selected algorithms on the training data to learn patterns and
relationships within the data.
• Hyperparameter Tuning: Optimizing model parameters to improve performance through
techniques like grid search or randomized search.
• Cross-Validation: Evaluating model performance using techniques like k-fold cross-
validation to ensure robustness and avoid overfitting.
• Model Evaluation: Assessing model performance metrics such as accuracy, precision,
recall, and F1-score on the testing set to gauge predictive capability.

5.2.7 Choose best model for prediction

• Performance Metrics: Consider performance metrics relevant to your problem, such as

accuracy, or mean squared error (MSE) for regression tasks.
• Compare Results: Compare the performance of different models based on the chosen
metrics. Identify the model that achieves the highest performance and generalizes well to
unseen data.

39
5.3 Implemented Algorithms:

5.3.1. Random Forest Random Forest is a popular machine learning algorithm that belongs
to the supervised learning technique. It can be used for both Classification and Regression
problems in ML.
• It is based on the concept of ensemble learning, which is a process of combining multiple
classifiers to solve a complex problem and to improve the performance of the model. As
the name suggests, Fig 5.1 shows "Random Forest is a regressor that contains a number
of decision trees on various subsets of the given dataset and takes the average to improve
the predictive accuracy, mae and rsme of that dataset."
• Instead of relying on one decision tree, the random forest takes the prediction from each
tree and based on the majority votes of predictions, and it predicts the final output. The
greater number of trees in the forest leads to higher accuracy and prevents the problem of
overfitting.

Random Forest Regressor

Model performance for Training set
Data Rows: 1610
- Root Mean Squared Error: 0.0238
- Mean Absolute Error: 0.0030
- R2 Score: 0.9998
----------------------------------
Model performance for Test set
Data Rows: 403
- Root Mean Squared Error: 0.1053
- Mean Absolute Error: 0.0155
- R2 Score: 0.9959

Fig 5.1 Random Forest

40
5.3.2. Support Vector Machine Support Vector Machines (SVMs) shown in Fig 5.2 are a
popular class of supervised machine learning algorithms used for classification and regression
analysis. They were originally developed in the 1990s by Vladimir Vapnik and his colleagues
and have since become widely used in many different fields. The basic idea behind SVMs is
to find a hyperplane in a high-dimensional space that separates different classes of data points.
The hyperplane is chosen in such a way that it maximizes the distance between the closest
data points from each class.
• These closest data points are called support vectors, and the distance between them is
known as the margin. The SVM algorithm aims to find the hyperplane with the maximum
margin, as this is likely to be the one that generalizes best to new data.
• In the case of non-linearly separable data, SVMs use a technique called kernelization,
where the data is mapped to a higher-dimensional space where a linear separation is
possible. This is achieved by applying a kernel function to the data, which computes the dot
product of the input vectors in the higher-dimensional space.

Support Vector Machine

Model performance for Training set
Data Rows: 1610
- Root Mean Squared Error: 0.4095
- Mean Absolute Error: 0.1242
- R2 Score: 0.9339
----------------------------------
Model performance for Test set
Data Rows: 403
- Root Mean Squared Error: 0.5547
- Mean Absolute Error: 0.2134
- R2 Score: 0.8874

Fig 5.2 Support Vector Machines

5.3.3. Decision Trees Decision trees recursively partition the dataset based on feature values.
At each step, the algorithm selects the feature that best splits the data, creating nodes
representing decisions based on these features shown in Fig 5.3. The process continues until a
stopping criterion is met, such as a maximum tree depth or no further improvement in
impurity. Finally, leaf nodes are created to make predictions based on the majority class or
average value of the samples in each node.

41
 Decision Tree
Model performance for Training set
Data Rows: 1610
- Root Mean Squared Error: 0.0000
- Mean Absolute Error: 0.0000
- R2 Score: 1.0000
----------------------------------
Model performance for Test set
Data Rows: 403
- Root Mean Squared Error: 0.1220
- Mean Absolute Error: 0.0149
- R2 Score: 0.9946

Fig 5.3 Decision Tree

5.3.4. K-nearest Neighbors (Knn) Regressor The KNN regressor algorithm shown in Fig 5.4
operates on the principle of similarity: it assumes that data points with similar feature values
tend to have similar target values. During prediction, the algorithm identifies the K nearest
neighbors to the new data point in the feature space and assigns a predicted target value based
on the average (or weighted average) of these neighbours target values. The choice of K
affects the model's bias-variance trade off: smaller values of K result in more flexible models
but may lead to overfitting, while larger values of K lead to smoother predictions but may
sacrifice model flexibility.

K-Neighbors Regressor
Model performance for Training set
Data Rows: 1610
- Root Mean Squared Error: 0.4280
- Mean Absolute Error: 0.2581
- R2 Score: 0.9278
----------------------------------
Model performance for Test set
Data Rows: 403
- Root Mean Squared Error: 0.5617
- Mean Absolute Error: 0.3648
- R2 Score: 0.8845

Fig 5.4 K-Nearest Neighbour

42
5.3.5 Linear Regression

Linear regression aims to model the relationship between the input features and the target
variable as shown in Fig 5.5 by fitting a linear equation to the observed data. The algorithm
estimates the coefficients of this equation during the training phase, typically using
optimization techniques such as Ordinary Least Squares (OLS) or Gradient Descent. The
resulting linear model provides a straightforward interpretation of the relationship between
each feature and the target variable. However, linear regression assumes a linear relationship
between the features and the target, which may not always hold true in practice. Additionally,
it is sensitive to outliers and multicollinearity among the input features. Regularization
techniques such as Ridge regression or Lasso regression can be employed to address these
issues and improve the robustness of the linear regression model.

Linear Regression
Model performance for Training set Data
Rows: 1610
- Root Mean Squared Error: 0.5369
- Mean Absolute Error: 0.4177
- R2 Score: 0.8863
----------------------------------
Model performance for Test set Data
Rows: 403
- Root Mean Squared Error: 0.5949
- Mean Absolute Error: 0.4568
- R2 Score: 0.8705

Fig 5.5 Linear Regression

43
5.4 Summary

The machine learning models were evaluated using various metrics including
R2 score, accuracy, root mean squared error (RMSE), and mean absolute error (MAE).
Among the models tested, the Random Forest Regressor demonstrated the highest
performance, achieving an impressive R2 score of 0.995854 and an accuracy of 99.59%, with
relatively low RMSE (10.64) and MAE (1.63) values. Following closely behind, the Decision
Tree model also exhibited excellent performance with a slightly lower R2 score of 0.994551
and accuracy of 99.46%, accompanied by RMSE (12.20) and MAE (1.49) metrics comparable
to the Random Forest Regressor. R2 score of 0.994551 and an accuracy of 99.46%. While its
RMSE (12.20) and MAE (1.49) metrics were slightly higher compared to the Random Forest
Regressor, they still reflected strong predictive capabilities. However, the Support Vector
Machine (SVM), K-Neighbors Regressor, and Linear Regression models displayed relatively
inferior performance metrics. Despite achieving moderate levels of accuracy (SVM: 88.74%,
K-Neighbors: 88.45%, Linear Regression: 87.05%).The most convenient algorithm having
good metrics is Random forest regressor.

44
CHAPTER-6

45
 6. RESULTS AND ANALYSIS
6.1. Introduction

Utilizing the Random Forest Regressor within a Flask application for student
performance analysis and prediction represents a robust approach to leveraging machine
learning for educational insights. The Random Forest's ability to capture complex
relationships within the dataset makes it well-suited for predicting student performance based
on various factors. The Flask application serves as an intuitive platform for users, such as
educators or administrators, to input relevant student data and receive predictions regarding
academic outcomes. Through the Flask application, users can interactively explore the factors
influencing student performance, such as demographic information, previous academic
achievements, extracurricular activities, and socio-economic backgrounds as shown in Fig 6.1.
By visualizing these factors and their impact on performance, the data is shown in the table
6.1 that can gain actionable insights into student behaviour and tailor interventions to support
struggling students or enhance the learning experience for all. Additionally, the Flask
application can facilitate longitudinal analysis by tracking student progress over time,
enabling educators to assess the effectiveness of interventions and refine strategies
accordingly. Fig 6.2 and Fig 6.3 show the eda analysis that say the Data visualization plays a
crucial role in the application by providing intuitive representations of student performance
trends, comparative analyses, and predictive insights. Visualizations such as scatter plots,
histograms, and bar charts can offer comprehensive views of student data, aiding educators in
making informed decisions about resource allocation, curriculum design, and personalized
learning pathways.

Overall, the integration of the Random Forest Regressor model with Flask-
based student performance analysis and prediction offers a powerful platform for enhancing
educational practices, promoting data-driven decision-making, and ultimately, improving
student outcomes. Student performance score is labelled for indicating student performance
that is based many attributes such as Attendance, Cgpa Skill credits, Branch, Section, Gender
etc. and those information indicates the performance of the students shown in Table 6.1.

46
6.2 Results

Table 6.1 Data

Fig 6.1 Correlation with Performance score

47
Year of joining counts Branch counts

Section Counts Gender Counts

Grade counts Student performance counts

Fig 6.2 Univariate analysis graphs

48
Fig 6.3 Multivariate analysis graphs

49
Table 6.2 Train data Results

Table 6.3 Test data Results

Fig 6.4 Model comparison

50
Fig 6.5 Login Page

Fig 6.6 Homepage

51
Fig 6.7 Summary Statistics

Fig 6.8 Visualization

52
Fig 6.9 Prediction Page

Fig 6.10 Prediction Result

53
6.3 Analysis of performance metrics

Upon meticulous examination of the model performance metrics on both the

training and test datasets, distinct performance trends emerge among the evaluated regression
models. The Random Forest Regressor stands out prominently, demonstrating unparalleled
predictive efficacy across various metrics. Notably, it achieves the lowest Root Mean Squared
Error (RMSE) and Mean Absolute Error (MAE) values alongside the highest R2 scores on
both sets. Specifically, on the training set, the Random Forest Regressor attains an
impressively low RMSE of 0.0238, MAE of 0.0030, and a remarkable R2 score of 0.9998.
Similarly, on the test set, its performance remains robust with an RMSE of 0.1053, MAE of
0.0155, and an outstanding R2 score of 0.9959. Conversely, while the Decision Tree model
showcases impeccable performance on the training set with zero RMSE and MAE and a
perfect R2 score of 1.0000, its inability to generalize becomes apparent on the test set, where
it exhibits higher error rates. The Support Vector Machine (SVM) demonstrates commendable
performance but falls short of the Random Forest Regressor in terms of accuracy and
generalization. The K-Neighbors Regressor and Linear Regression models also present
relatively higher errors, suggesting suboptimal predictive capabilities compared to the
Random Forest shown comparatively in Fig 6.4 from the test and train data results as shown in
Table 6.2 and Table 6.3.and the Model is selected for prediction in which the flask interface is
used for running application as shown in Fig 6.5 for login and Fig 6.6 for home page. Data
Visualization done in Data visualization dashboard as shown in Fig 6.7 and Fig 6.8 and
Prediction page shows the prediction results as shown in Fig 6.9 and Fig 6.10.

Overall, the exhaustive consideration of these performance metrics underscores

the Random Forest Regressor's unparalleled prowess in achieving precise and robust
predictions across both training and test datasets. The Support Vector Machine (SVM) exhibits
relatively higher errors on both training (RMSE of 0.4095, MAE of 0.1242, R2 score of
0.9339) and test (RMSE of 0.5547, MAE of 0.2134, R2 score of 0.8874) sets compared to the
Random Forest. While the K-Neighbors Regressor and Linear Regression models also
perform reasonably well, their higher errors indicate suboptimal predictive capabilities
compared to the Random Forest Regressor. Thus, based on these detailed performance values,
the Random Forest Regressor emerges as the preferred choice for accurate prediction tasks.

54
6.4 Discussion of Findings

The initial stages of the machine learning project involve understanding the
problem statement and collecting the necessary data. In this case, the project aims to predict
student performance based on various factors such as branch, section, gender, attendance,
GPA, grades, and skills. The dataset, named "projectdata.csv," comprises 12 columns and
2012 rows. After importing the required packages and loading the dataset into a Pandas Data
Frame, initial exploratory steps are taken. Data checks are performed to ensure data quality.
These include checking for missing values, duplicates, data types, the number of unique
values in each column, and basic statistics of the dataset. Fortunately, there are no missing
values or duplicates, and the data types seem appropriate for further analysis. The dataset
contains a total of 2013 entries with 12 columns, including numerical and categorical features.

Exploratory data analysis (EDA) is conducted to gain insights into the

distribution and relationships between different variables. Univariate analysis examines
individual attributes, such as the distribution of students across joining years, branches,
sections, and genders. Bivariate and multivariate analyses explore relationships between
variables, such as the correlation between attendance, GPA, and performance score. Data
preprocessing steps involve data cleansing, reduction, transformation, and enrichment. In this
case, certain columns like "Roll No," "Email Id," "Student Performance," and other encoded
categorical variables are dropped to focus on the relevant features for model training. This
step streamlines the dataset for further analysis and model development.

Overall, the project follows a structured approach, starting from understanding

the problem statement, collecting data, performing data checks and exploratory analysis, to
preprocessing the data for model training. These steps lay the foundation for building
predictive models to understand the factors influencing student performance.This entails an
array of tasks, including but not limited to handling missing values, identifying and addressing
duplicates, verifying data types, and assessing the distribution of unique values across each
column. The "projectdata.csv" dataset, comprising 2012 rows and 12 columns, exhibits no
conspicuous anomalies such as missing values or duplicates. Moreover, categorical variables
may be encoded to enhance their interpretability and utility in subsequent analyses.

55
 Integrating a predictive model into a Flask application for data visualization
and prediction is a sophisticated approach that enhances the usability and accessibility of the
insights derived from the model. Flask provides a flexible framework for building web
applications, allowing you to create interactive interfaces that cater to the specific needs of
your users. By leveraging Flask's capabilities, you can streamline the process of data input,
analysis, and prediction, all within a user-friendly web environment. One key aspect of this
integration is the visualization of data. With Flask, you can incorporate popular visualization
libraries such as Matplotlib to create dynamic and informative visualizations. These
visualizations can help users gain a deeper understanding of the underlying data by
highlighting trends, patterns, and relationships. By presenting data in a visual format, you
make it more accessible and engaging, enabling users to explore and interpret the data more
effectively.
In addition to data visualization, Flask allows you to seamlessly integrate
predictive modelling capabilities into your application. This involves creating a form or
interface where users can input data for which they want predictions. Once the user submits
the input data, Flask passes it to the predictive model, which generates predictions based on
the provided features. The predictions can then be displayed to the user, either alongside the
input data or in a separate section of the application. Furthermore, incorporating a feedback
mechanism into the application allows users to provide input on the accuracy and relevance of
the predictions. This feedback loop is invaluable for iteratively improving the predictive
model and enhancing its performance over time. By soliciting user feedback, you can refine
the model based on real-world insights and ensure that it remains relevant and effective in
addressing the needs of its users. Lastly, scalability is a crucial consideration when developing
a Flask application for data visualization and prediction. As the application gains users and
processes larger datasets, it must be able to handle increased traffic and data processing
demands efficiently.

Integrating a predictive model into a Flask application for data visualization

and prediction offers a powerful solution for empowering users to make informed decisions
based on data-driven insights. By leveraging Flask's capabilities for web development
alongside data visualization and predictive modelling techniques, you can create a versatile
and user-friendly application that meets the needs of its users effectively.

56
6.5 Summary

Integrating predictive models into Flask applications for data visualization and
prediction offers a powerful solution for delivering data-driven insights to users. Flask's
flexibility enables the creation of interactive interfaces that facilitate seamless data input,
analysis, and prediction. By incorporating popular visualization libraries like Matplotlib or,
developers can create dynamic visualizations that enhance user understanding of the
underlying data, highlighting trends and patterns. In addition to visualization, Flask facilitates
the integration of predictive modelling capabilities, allowing users to input data for prediction
through a user-friendly interface. Once submitted, Flask passes this data to the predictive
model, which generates predictions based on provided features. Once submitted, Flask
forwards this data to the predictive model, which generates predictions based on the provided
features. Furthermore, incorporating feedback mechanisms allows users to provide input on
prediction accuracy, enabling iterative improvements to the model over time. Flask
applications for data visualization and prediction offer a versatile solution for empowering
users with data-driven insights. By leveraging Flask's capabilities alongside visualization and
predictive modelling techniques, developers can create user-friendly applications that meet the
evolving needs of their users effectively.

57
CHAPTER-7

58
 7. CONCLUSION AND FUTURE ENHANCEMENTS

7.1 Conclusion

In conclusion, the application of machine learning in the realm of student

performance analysis and prediction, specifically utilizing the Random Forest Regressor
model, marks a significant advancement in education. By employing this model, educational
stakeholders can obtain precise predictions of student performance scores based on a
multitude of input features. Leveraging features such as grades, attendance records, socio-
demographic data, and more, the Random Forest Regressor facilitates a comprehensive
understanding of factors influencing student outcomes. Moreover, the incorporation of
exploratory data analysis (EDA) techniques enriches the predictive process through insightful
data visualization. EDA enables stakeholders to glean actionable insights from the data,
identify patterns, and discern correlations between various input features and student
performance. This visual representation enhances the interpretability and transparency of the
predictive model's outcomes, fostering informed decision-making among educators,
administrators, and policymakers. The integration of a Flask application further streamlines
the utilization of these predictive capabilities, offering a user-friendly interface for accessing
and interpreting student performance predictions. This web-based platform enhances
accessibility, enabling educators to easily access predictive insights and make timely
interventions to support students at risk of falling behind.

However, amidst these advancements, it is imperative to acknowledge and

address ethical considerations inherent in the deployment of machine learning in education.
Safeguarding student privacy, mitigating biases in the data or models, and ensuring
transparency and accountability in the predictive process are paramount. In summary, the
fusion of machine learning techniques, including the Random Forest Regressor model,
exploratory data analysis, and a Flask application, presents a powerful approach to student
performance analysis and prediction. This holistic framework empowers educational
stakeholders with actionable insights to optimize learning outcomes, foster student success,
and enhance the overall educational experience.

59
7.2 Challenges Encountered and Limitations

7.2.1 Challenges

• Data Quality and Availability: Ensuring high-quality and accessible data in educational
settings presents a significant challenge due to issues like incompleteness, inconsistency,
and errors, which can undermine the accuracy and reliability of predictive models.

• Feature Selection and Engineering: Identifying relevant features and crafting

informative predictors is crucial for model effectiveness. However, determining the most
influential features and transforming them appropriately can be daunting, especially with
diverse datasets.

• Overfitting and Generalization: Preventing overfitting and ensuring adequate

generalization of models, especially Random Forests, to new data is a balancing act.
Overfitting, particularly on limited or noisy data, poses a significant challenge.

7.2.1 Limitations

• Model Interpretability: Despite their high prediction accuracy, Random Forest Regressor
models can be complex and challenging to interpret. Understanding model decisions and
communicating them effectively to stakeholders may pose limitations.

• Bias and Fairness: Biases present in data or introduced by models can lead to unfair
predictions, limiting model effectiveness. Mitigating biases and ensuring fairness in
predictive models is essential for ethical and equitable decision-making.

• Ethical Considerations: Deploying machine learning models in educational settings raises

ethical concerns regarding privacy, transparency, and accountability. Ensuring compliance
with ethical guidelines and addressing concerns about model transparency and
accountability are limitations that need to be addressed.

60
7.3 Future Enhancements

Future enhancements for student analysis and prediction using machine learning could focus
on several key areas to improve the accuracy, interpretability, and impact of the models. These
enhancements could include:

• Incorporating Advanced Machine Learning Techniques: Future models could

leverage advanced machine learning techniques such as deep learning and ensemble
methods to improve prediction accuracy. Deep learning models, such as neural networks,
can capture complex patterns in the data, while ensemble methods can combine multiple
models to enhance performance.

• Enhancing Model Interpretability: Improving the interpretability of machine learning

models is crucial for gaining insights into the factors influencing student performance.

• Personalized Learning Recommendations: Future models could provide personalized

learning recommendations based on individual student profiles, helping educators tailor
their teaching strategies to meet the specific needs of each student.

• Real-time Monitoring and Intervention: Implementing real-time monitoring of student

performance could enable early intervention strategies for students at risk of academic
failure. By identifying struggling students early, educators can provide timely support to
improve outcomes.

• Integration with Learning Management Systems (LMS): Integrating the predictive

models with LMS could provide a seamless experience for educators, allowing them to
access student performance predictions directly within their existing workflow.

Overall, future enhancements should aim to improve the accuracy, interpretability, and ethical
considerations of machine learning models for student analysis and prediction, ultimately
leading to better outcomes for students and educators alike.

61
CHAPTER-8

62
 8. REFERENCES

[1] D.Solomon, S. Patil, and P. Agrawal, Predicting performance and poten tial dif culties of
university student using classi cation: Survey paper, Int. J. Pure Appl. Math, vol. 118, no. 18,
pp. 27032707, 2018.

[2] E. Alyahyan and D. Dütegör, Predicting academic success in higher education: Literature
review and best practices, Int. J. Educ. Technol. Higher Educ., vol. 17, no. 1, Dec. 2020.

[3] V. L. Miguéis, A. Freitas, P. J. V. Garcia, and A. Silva, Early segmen tation of students
according to their academic performance: A predictive modelling approach,
Decis.SupportSyst.,vol.115,pp. 3651,Nov.2018.

[4] P. M. Moreno-Marcos, T.-C. Pong, P. J. Munoz-Merino, and C. D. Kloos, Analysis of the

factors in uencing Learners performance prediction with learning analytics, IEEE Access,
vol. 8, pp. 52645282, 2020.

[5] A. E. Tatar and D. Dütegör, Prediction of academic performance at undergraduate

graduation: Course grades or grade point average? Appl. Sci., vol. 10, no. 14, pp. 115, 2020.

[6] Y.Zhang, Y.Yun, H.Dai, J.Cui, and X.Shang, Graphs regularized robust matrix
factorization and its application on student grade prediction, Appl. Sci., vol. 10, p. 1755, Jan.
2020.

[7] H.Aldowah, H.Al-Samarraie, and W.M.Fauzy, Educational datamining and learning

analytics for 21st century higher education: A review and synthesis, Telematics Informat.,
vol. 37, pp. 1349, Apr. 2019.

[8] K. L.-M. Ang, F. L. Ge, and K. P. Seng, Big educational data & analytics: Survey,
architecture and challenges, IEEE Access, vol. 8, pp. 116392116414, 2020.

[9] A. Hellas, P. Ihantola, A. Petersen, V. V. Ajanovski, M. Gutica, T. Hynninen, A.Knutas, &

63
J.Leinonen, C.Messom, and S.N. Liao, Predict ing academic performance: A systematic
literature review, in Proc. 23rd Annu. Conf. Innov. Technol. Comput. Sci. Educ., Jul. 2018,
pp. 175199.

[10] L. M. Abu Zohair, Prediction of students performance by modelling small dataset size,
Int. J. Educ. Technol. Higher Educ., vol. 16, no. 1, pp. 18, Dec. 2019, doi: 10.1186/s41239-
019-0160-3.

[11] X. Zhang, R. Xue, B. Liu, W. Lu, and Y. Zhang, Grade prediction of student academic
performance with multiple classification models, in Proc. 14th Int. Conf. Natural Comput.,
Fuzzy Syst. Knowl. Discovery (ICNC-FSKD), Jul. 2018, pp. 10861090.

[12] S. T. Jishan, R. I. Rashu, N. Haque, and R. M. Rahman, Improving accuracy of students

nal grade prediction model using optimal equal width binning and synthetic minority over-
sampling technique, Decis. Anal., vol. 2, no. 1, pp. 125, Dec. 2015.

[13] A. Polyzou and G. Karypis, Grade prediction with models speci c to students and
courses, Int. J. DataSci. Anal., vol. 2, nos. 34, pp. 159171, Dec. 2016.

[14] Z. Iqbal, J. Qadir, A. N. Mian, and F. Kamiran, Machine learning based student grade
prediction: A case study, 2017, arXiv:1708.08744. [Online]. Available:
https://arxiv.org/abs/1708.08744

[15] I. Khan, A. Al Sadiri, A. R. Ahmad, and N. Jabeur, Tracking student performance in

introductory programming by Meansof machine learning, in Proc. 4th MEC Int. Conf. Big
Data Smart City (ICBDSC), Jan. 2019, pp. 16.

64
CHAPTER-9

65
 9. APPENDICES

User Manual for Student Performance Prediction Application

1. Introduction

2. Getting Started

- Accessing the Application

- User Registration and Login

3. Predicting Student Performance

- Input Features

- Making Predictions

4. Interpreting Predictions

- Understanding Prediction Outputs

5. Data Visualization and Summary Statistics

- Exploring Input Features

- Visualizing Attributes

6. Viewing Prediction History

- Accessing Past Predictions in Data

1. Introduction

Welcome to the Student Performance Prediction Application! This user manual provides
detailed instructions on how to use the application to predict student performance based
on various input features. The application utilizes machine learning techniques,
specifically the Random Forest Regressor model, to generate predictions and assist
educators in supporting student success.

66
2. Getting Started

• Accessing the Application

To access the application, run the per-predictionapp.py. And visit localhost address:
http://127.0.0.1:5000/ arrival, you will be greeted with the application's landing page, where
you can register for a new account or log in if you already have an existing account.

• User Registration and Login

If you are a new user, click on the "Register" button and fill out the registration form with
your details. After registration, you will receive a confirmation email to verify your account.
Once verified, you can log in using your credentials.

3. Predicting Student Performance

• Input Features

Before making predictions, you will need to input relevant student data into the application.
The input features may include:

- Student demographics (e.g., age, gender)

- Academic records (e.g., CGPA, Attendance)

- Socio-economic background

Ensure that you have the necessary data available before proceeding with the prediction
process.

• Making Predictions

Once logged in, navigate to the "Prediction" tab in the application menu. Here, you will find a
form where you can input student data. Fill out the required fields with the appropriate
information and click on the "Predict" button.

67
The application will then utilize the Random Forest Regressor model to generate predictions
of student performance based on the input data. The prediction results will be displayed on
the screen, providing insights into the expected performance score.

4. Interpreting Predictions

• Understanding Prediction Outputs

The prediction outputs will include the predicted student performance score, along with any
additional insights provided by the model. Interpret the prediction results to understand the
anticipated performance level of the student based on the input data provided.

5. Data Visualization and Summary Statistics

• Exploring Input Features

Navigate to the "Data Visualization" tab in the application menu to explore summary
statistics and visualize input features. This section provides insights into the distribution,
variability, and relationships between different attributes of the student data.

• Visualizing Attributes

Use the interactive visualization tools to explore individual attributes and their correlations
with student performance. Visual representations such as histograms, scatter plots, and box
plots help identify patterns and trends in the data, facilitating deeper understanding and
analysis.

6. Viewing Prediction History

Accessing Past Predictions

To view past predictions, navigate to the “projectdata.csv“ file to view all data along with
predicted data at last on a new student data in the project file. Here, you will find a record of
all previous predictions made using the application.

User manual provides comprehensive instructions for users to navigate the Flask application,
input student data, make predictions, interpret prediction outputs, explore input features
through data visualization, and view past prediction history.

68