Applied Mathematics and Nonlinear Sciences, 9(1) (2024) 1-21

Applied Mathematics and Nonlinear Sciences

https://www.sciendo.com

Convergence of Artificial Intelligence Algorithms and Education Fuels Educational

Reform in Higher Education Programs

Jingjing Nie1,†
1. School of Mechanical and Electrical Engineering, Wuhan Business University, Wuhan, Hubei,
430000, China.

Submission Info

Communicated by Z. Sabir
Received February 15, 2024
Accepted April 27, 2024
Available online May 30, 2024

Abstract
In the realm of higher education management, data mining has emerged as a potent tool for elevating the quality of
instruction and administrative operations. This study presents the development of a curriculum validation analysis model
utilizing association rule clustering. We collected and processed data from design students at a university, addressing
missing values and eliminating outliers. Through data mining, we explored the relationships between students’
coursework, skill acquisition, and the demands of potential employers in society. This analysis underpinned a series of
reforms in higher education curricula. To assess the impact of these curricular changes, two instructors conducted both
process and outcome evaluations of the revised programs. The evaluations revealed that five student groups achieved
scores above 80, one group scored between 70 and 80, and three groups scored between 60 and 70. All student groups
scored above 60, demonstrating their competence in fundamental job-related tasks and validating their readiness for
professional employment. This research offers valuable insights for the revision of professional training programs and
curriculum reform in higher education.

Keywords: Association rule clustering; Similarity measures; Systematic clustering method; Curriculum reforms.
AMS 2010 codes: 97P10

†Corresponding author.
Email address: [email protected] ISSN 2444-8656
https://doi.org/10.2478/amns-2024-1267
© 2023 Jingjing Nie, published by Sciendo.
This work is licensed under the Creative Commons Attribution alone 4.0 License.
 2 Jingjing Nie. Applied Mathematics and Nonlinear Sciences, 9(1) (2024) 1-21

1 Introduction

Artificial intelligence, as a new round of strategic technology leading the future development and the
core driving force of the new round of industrial change, is integrating with various fields at a very
fast development speed [1-2]. To promote the construction of new engineering and new agricultural
science in the field of artificial intelligence in colleges and universities shows that in the face of the
characteristics of the new era, the cultivation of innovative talents with computational thinking and
the ability to flexibly apply the combination of artificial intelligence technology needs to be carried
out throughout the various stages of education, which has become an urgent task of education in the
current era [3-5]. The construction of new engineering new agricultural science should be combined
with the development of the artificial intelligence era to cultivate innovative consciousness, adapt to
and lead the future development of talents, promote the reform of higher education of the new strategy,
the formation of “artificial intelligence + X” to cultivate composite professionals of the new model
[6]. More and more majors have introduced AI courses and other related courses into the training
system, tried to combine AI courses with the content of the field of their majors, and reformed and
explored the teaching mode of the courses [7].

In the current wave of education and teaching reform, the contradiction between “large-scale
education” and “individualized training” is still prominent. Ancient education generally used the
“individualized” teaching method, which can easily meet the needs of the implementation of tailor-
made education [8-9]. In modern education, however, due to the high demand for higher education
rates, the teaching mode is too single, the teaching evaluation method is insufficient, the learning
burden of students is slightly heavy, and other factors, it is difficult to grasp the physical and mental
characteristics of each student and the differences between students. It isn’t easy to ensure that
different students can enjoy personalized education [10]. Therefore, artificial intelligence to
accelerate the cultivation of diversified composite talents to help the scenario of tailor-made education
landing is the new mission of education innovation and development in the era of intelligence [11].

The development of artificial intelligence not only promotes technological change in various fields
but also promotes the change of the education system. Promoting the cross-fertilization of disciplines
and emphasizing the combination of AI professional fields has become a new idea of discipline
construction, and more and more scholars have noticed the importance of the development of
professional fields based on AI technology and have explored it in the practice of education and
teaching [12]. Literature [13] explored the status quo of AI technology applied to education in
countries around the world and found that the research in AI education-related fields has steadily
increased, and summarized the important research classroom direction of AI education, which has a
positive significance for the development and study of AI education. Literature [14] combines fuzzy
affiliation function and convolutional neural network methods to detect and predict students’
classroom performance, which can realize the effective detection of classroom plagiarism. Literature
[15] proposes the application of artificial intelligence technology to personalized learning, which is
used to differentiate students’ personalities and learning preferences in order to provide customized
learning solutions with a higher degree of fitness, which helps to improve students’ performance.
Literature [16] developed an interdisciplinary analytical framework to ensure effective feedback on
student learning in AI digital learning models and validated the effectiveness of the proposed
analytical framework through teaching practice. Literature [17] describes and enumerates the changes
brought about by the application of AI technology in the field of education, which effectively assists
teachers in their teaching management and relieves them of repetitive and tedious work.

The rapid development of a new generation of intelligent technologies such as artificial intelligence,
knowledge graphs, and personalized recommendations, and the continuous deepening of its teaching
has pushed educational change and innovation to a deeper depth and injected new kinetic energy into
 Convergence of Artificial Intelligence Algorithms and Education Fuels Educational Reform in Higher 3
Education Programs

the implementation of the scale and personalization of teaching according to the talent in the
intelligent era. Literature [18] reveals that programming languages and AI technology-enriched
teaching models provide students with an easy, convenient teaching experience but are unable to
effectively support students in tackling complex learning tasks. Literature [19] explored the
possibility of personalized teaching models based on AI technology perspectives. It illuminated the
specific elements that need to be focused on, i.e., learning motivation, instructional diversity, and
algorithmic bias. Literature [20] points out that Benner’s growth theory of the individual’s pursuit of
supreme harmony, as well as the interaction between human beings and the world is the essence of
human development, lays a theoretical foundation for educational research and proposes to look at
educational reforms with a critical eye. Literature [21] describes the collective agency of teachers in
the context of curricular education reform, which builds a platform of a good interactive atmosphere
for the professional development, communication, and collaboration of individual teachers and helps
the professional development of teachers in the context of the era of curricular reform. Literature [22]
emphasizes the important role of education-related artifacts in promoting education reform and
education policy-making and discusses it with the case of physical education curriculum reform.

In this paper, association rule clustering is first introduced as the basis for constructing a model for
validating course curricula. Then, the model is used to analyze the data of students majoring in
design in a college, to explore the relevance of the college curriculum, to cluster the association
rules on the school learning data and employer’s feedback data, and to optimize the curriculum
according to the frequent association rules obtained. Finally, the teaching effect of the new
curriculum is evaluated in process and result, and the learning outcomes of the core professional
courses and vocational skills courses of the nine groups of students are examined so as to verify the
effect of student cultivation after the curriculum reform and the validity of the analytical model in
this paper.

2 Method

2.1 Association rule clustering

To determine whether the curriculum system is reasonable or not, you can find the rules from the data,
especially the feedback from employer enterprises and the performance of students during their
school years, etc. They are very suitable for conducting correlation rule analysis.

It’s not an easy task to find rules that are worthy of attention from a large number of people. Although
the number of rules can be bounded using interest metrics such as Lift to identify potentially valuable
rules, whether a rule is noteworthy or not is largely subjective, and it depends on the user’s expertise
and current analysis goals. In addition to interest metrics, another technique for mining valuable
association rules is the visualization of association rules. This visualization method focuses on
displaying the internal information of the association rules so that the rules can be arranged in some
reasonable way to discover content of interest. Displaying too many rules in the visualization space
can greatly diminish the usefulness of revealing the internal information of the rules.

The association rules are analyzed and reorganized using cluster analysis in this study. A new method
of association rule similarity metric is proposed: similarity weighted metric, which analyzes the
similarity of rules in terms of both content and structure of association rules and calculates the
distance between the rules so as to fully present the overall structural information of association rules,
which is conducive to the researcher’s discovery of the information of interest.
 4 Jingjing Nie. Applied Mathematics and Nonlinear Sciences, 9(1) (2024) 1-21

Clustering association rules require calculating the similarity between the rules. The characteristics
of the association define the similarity between rules themselves, such as support, confidence,
usefulness, etc. Such a metric for determining the similarity of association rules has some defects and
does not provide a comprehensive measure of the similarity between association rules. In order to
measure the similarity between two association rules more comprehensively, it is necessary to analyze
the source of the differences between the association rules. Structural differences (the set of
antecedent items and the set of consequent items) and content differences (all items contained in the
association rule) are the two types of differences between association rules. Therefore, we propose a
new method of measuring similarity called similarity weighted measure.

Definition 1 Let I = I1 , I 2 ,, I m  be the set of all terms and the sets of terms T1 and T2 be non-
empty subsets of I , i.e., T1  I , T2  I , and T1   , T2   , then the similarity between the sets of
terms T1 and T2 is defined:

T1  T2
Sim (T1 , T2 ) = (1)
T1  T2

Definition 2 has two association rules:

R1 : X 1  Y1
(2)
R2 : X 2  Y2

The structural similarity of rules R1 and R2 is defined:

RSim ( R1 , R2 ) =   Sim ( X 1 , X 2 ) +   Sim (Y1 , Y2 )

(3)
  0,   0,  +  = 1

Definition 3 has two association rules:

R1 : X 1  Y1
(4)
R2 : X 2  Y2

The content similarity of rules R1 and R2 is defined:

s ( R1  R2 )
CSim ( R1 , R2 ) =
s ( R1 ) + s ( R2 ) − s ( R1  R2 )
(5)

Definition 4 has two association rules

R1 : X 1  Y1
(6)
R2 : X 2  Y2

The similarity of rules R1 and R2 is defined:

 Convergence of Artificial Intelligence Algorithms and Education Fuels Educational Reform in Higher 5
Education Programs

Sim ( R1 , R2 ) =  1  RSim ( R1 , R2 ) +  2  CSim ( R1 , R2 )

(7)
 1  0,  2  0,  1 +  2 = 1

In this definition, the similarity of a rule is measured by two parts, i.e., The structure and content of
the rule are used to calculate the similarity measure of the rule. The structure of the rules and the
content of the rules are equally important to the similarity measure between the rules, so
 1 = 0.5,  2 = 0.5 it is generally taken.

The similarity measure Sim( x, y ) can always be converted into the distance measure d ( x, y ) . When
converting the similarity measure into a formal distance measure, the distance measure must satisfy
some of the following properties:

1) Symmetry, i.e., for any x, y , there is a d ( x, y ) = d ( y, x) .

2) Non-negativity, i.e., for any x, y , there is d ( x, y )  0 . Also, d ( x, y ) = 0 if and only if x = y .

3) Satisfies the triangle inequality, i.e., for any x, y, z , it satisfies d ( x, y )  d ( x, z ) + d ( z, y ) .

Therefore, the association rule similarity measure Sim ( R1 , R2 ) translates into the definition of
the distance measure d ( R1 , R2 ) (as in Definition 5), a definition that is satisfied by the four
properties of the distance measure.

Definition 5 has two association rules

R1 : X 1  Y1
(8)
R2 : X 2  Y2

1 − Sim ( R1 , R2 )
d ( R1 , R2 ) = (9)
Sim ( R1 , R2 )

Asymmetric distance matrix consisting of association rule distances is as follows:

 0 d ( R1 , R2 ) L d ( R1 , Rn ) 
 
 d ( R2 , R1 ) 0 
D= (10)
 L 0 
 
 d ( Rn , R1 ) L 0 

The systematic clustering method, also often referred to as spectral clustering or hierarchical
clustering, is one of the most widely used methods today. The basic idea is to first consider n samples
as n classes (i.e., one class includes one sample) and then specify the distance between the samples
and the distance between the class and the class designation. The closest two classes are merged into
a new class, the distance between the new class and the other classes is calculated, and then the closest
two classes are merged, and so on, and finally, all the samples are in one class. The above merging
process is drawn into a clustering spectrum diagram. You can decide how many classes and what
samples each class has. The systematic clustering method involves the following:

1) Each sample forms a class that forms n a class.

 6 Jingjing Nie. Applied Mathematics and Nonlinear Sciences, 9(1) (2024) 1-21

2) Calculate the distance {d ( x, y )} between two two samples, noting D = (d ( x, y )) .

3) Merge the 2 classes with the closest distance into 1 new class.

4) Calculate the distance between the new class and the current class; if the number of classes is
equal to 1, then end the clustering process and draw a clustering spectrogram; otherwise, turn
to (3).

5) Determine the number of classes and the class. The class average method used in systematic
clustering takes full advantage of each sample’s information and is more representative than
other methods. Because the class-averaging method clusters better and is easy to understand
and calculate, it is widely used in scientific research. The method of calculating the distance
between classes by class averaging is as follows:

D ( G p , Gq ) =
1
  d ( x, y )
lm xG p yGq
(11)

Determining the number of clusters in a data set is a fundamental challenge in cluster analysis, for
which no satisfactory method can be found. A few commonly used statistics for determining the
number of classes are given below.

1) R 2 statistic

R2 = 1 −  Di / TSS (12)

To determine the number of n samples to be categorized based on the value of R 2 , firstly, the value
of R 2 for the most suitable categorization should not be too small, and it is better to reach 0.7 or
above. Secondly, the size of the merged R 2 value cannot be looked at in isolation, but the change in
the R 2 value should be looked at. The merging of classes always makes the R 2 value smaller, and
if this change is uniform, the merging is desirable. Still, when a certain merging makes the R 2 value
decrease a lot, this merging is not desirable, i.e., we have found the most suitable number of
classifications.

2) Half-bias R 2 statistic

When class G p and class Gq are merged into class Gm at the next level, the half-skewed R 2 statistic
is used to evaluate the effect of this merger, and the half-skewed R 2 formula is calculated:

Half biased R = Dw ( p, q ) / T
2
(13)

3) Pseudo F statistic PSF

The definition of the pseudo F statistic is similar to the definition of the F statistic in one-way
ANOVA:

Pseudo F (v(G − 1), v(n − G )) =

(TSS −  D ) / (G − 1)
i
(14)
 D / (n − G )
i
 Convergence of Artificial Intelligence Algorithms and Education Fuels Educational Reform in Higher 7
Education Programs

2
4) Pseudo t statistic PFT2

Dm − D p − Dq
Pseudo t (v, v( p + q − 2)) =
2

(D p + Dq ) / ( p + q − 2)
(15)

2
It is common to combine the pseudo F statistic with the pseudo t statistic, choosing the number of
categories such that the pseudo F statistic reaches a local peak. In contrast, the corresponding pseudo
t 2 statistic is smaller but larger at the next merging step.

5) Cubic Clustering Criterion (CCC)

Cubic Clustering Criterion CCC can be used for original hypothesis testing and estimating the number
of overall classifications.CCC is only applicable to coordinate data and cannot be used if the type of
input dataset is DISTANCE. Large CCC values indicate good clustering, and large negative values
indicate outliers.
2
The three best criteria for determining the number of classes are the pseudo F statistic. Pseudo t
statistic. Cubic clustering criterion CCC.

2.2 Educational reform of higher courses based on association rule clustering

The reform of higher education has focused on curriculum, but there is no unified definition of
curriculum as different researchers have different fields and perspectives. Curriculum can be divided
into five categories: subject theory, program theory, experience theory, expectation theory, and
teaching material theory.

Among them, the “subject theory” understands curriculum as teaching subjects and believes that the
subjects that appear in concrete form and are used for teaching in educational practice according to
the way of classification and logical order within the field are curriculum. In the broadest sense, it is
defined as the entire curriculum that students pursue, but in the narrowest sense, it is defined as a
specific course of study. The curriculum in this study is based on the “subject theory,” and the
curriculum is the subject of instruction.

Correlation in statistics, on the other hand, is when there is a correlation between two things, and one
changes because of the other. This trend of change can be divided into two types: one is the trend of
change between the two variables is the same; that is, A variable and B variable are in the trend of
growth or decline; such a trend is called positive correlation. The opposite trend of change is a
negative correlation. Positive and negative correlations are subdivided into correlations.

In the existing higher curriculum, due to the interconnection between the contents of various courses,
course learning outcomes, and practical work, the content and arrangement of a course are subject to
a variety of influences such as category, purpose, requirements, content, credit hours, etc. And this
interdependent and interconnected influence is called course relevance. Generally speaking, course
relevance includes the following two dimensions: one is the content dimension, i.e., there is a certain
correlation and connection between the contents of different courses; the other is the time dimension,
i.e., The relevance of different classes at their opening time and the correlation between the time when
the same thematic content appears in various courses.
 8 Jingjing Nie. Applied Mathematics and Nonlinear Sciences, 9(1) (2024) 1-21

By analyzing the relevance of university curriculum through data mining, it is possible to obtain the
correlation of various factors, such as the status and nature of the course, the category to which it
belongs, the situation of the class hours, the common software tools of the enterprise (which are
related to the course set up), the degree of mastery of the knowledge, and the expectations of the
enterprise on the skills, and so on, so as to provide guidelines for the reform of the curriculum.

2.2.1 Classification of courses

Teaching courses are divided into compulsory and elective courses. Since elective courses focus on
students’ interests and involve a variety of content, they have distinctive individual differences.
Therefore, this study mainly considers course content related to compulsory courses.

1) By course type

Generally speaking, general education courses are for all college students, and they are more
concerned with helping students participate in social life as members of society than with the purpose
of specialized courses. Core courses that develop students’ specialized knowledge and skills are
known as professional core courses.

Taking the actual course structure and the latest undergraduate talent cultivation program of
Educational Technology major in a university as an example, the discipline-specific professional
courses mainly include discipline basic courses, professional core courses, and characteristic
innovative courses, and the specific course classification is shown in Figure 1. It is evident that
Introduction to Educational Technology, Computer Fundamentals, etc. These are foundational
courses in the discipline, focusing on cultivating the level of basic theory. The specialized core
courses include sound media, educational television, multimedia courseware development, three-
dimensional animation creation, etc., which are related to the actual business and problems that will
be faced in the work.
 Convergence of Artificial Intelligence Algorithms and Education Fuels Educational Reform in Higher 9
Education Programs

Photographic techniques

Graphic animation
production
Introduction to
Object Oriented Educational Technology
Programming
Fundamentals of
Graphic Design Computer Applications

University Mathematics
Sound media technology Disciplinary (II) 1
Basic
Educational Television Courses University Mathematics
System (II) 2
Educational Web Fundamentals of
Development and Programming
Application

Educational Television Computerized Art

Programming and
Production Professional By course
core courses type
Educational Database Principles
Communication
C language and the
Multimedia courseware
teaching of information
design and development
technology in secondary
Distance education schools
applications Featured XML technology
Innovation
Digital video technology Classes Educational Game
Development
3D animation creation
Network Administration
Educational Technology and Maintenance
Research Methodology
Visualization
Instructional system Programming
design

Information Technology
Curriculum Pedagogy

Figure 1. Classification of courses by course type

2) By course direction

For students of different career directions, the curriculum groups of their cultivation directions and
the main focus of professional abilities are different. In order to make students more clear about the
professional competence they need to have, we at this moment discuss the correlation between the
courses of different directions to provide a scientific basis for the establishment of a curriculum
system and cultivation of talents in different directions, and at the same time, to provide reference
opinions and targeted guidance for the students to take direction courses according to their situation.

Still taking the educational technology major of a school as an example, the courses can be newly
classified into three professional directions. Combined with the actual curriculum arrangement, the
courses can be categorized by direction, and the specific course classification is shown in Figure 2.
In the new categorization method, the classes are divided into three content directions: education and
information technology, software engineering, and digital media courses used for teaching.
 10 Jingjing Nie. Applied Mathematics and Nonlinear Sciences, 9(1) (2024) 1-21

Photographic techniques

Computerized Art

Graphic animation
Writing skills training production

Teachers' oral training Graphic Design

Educational Psychology Sound media technology

Digital
Fundamentals of Education educational Educational Television
media System
Counseling for secondary Educational
school students Communication
C language and the Educational Television
teaching of information Programming and
technology in secondary Production
schools

Teacher Ethics and Digital video technology

Professional Development
3D animation creation
Study on Information
Technology Curriculum Information
By course
Standards and Teaching technology
direction
Materials education
Fundamentals of
Primary and Secondary Programming
Informatics Competitions Object Oriented
Comprehensive training in Programming
teaching competencies Educational Web
(microforms)1 Development and
Teaching design for Application
secondary school
information technology Database Principles
classrooms
Educational XML technology
Comprehensive training in software
teaching competencies engineering Multimedia courseware
(microforms)2 design and development

Instructional system design Distance education

applications
Information Technology
Network Administration
Curriculum Pedagogy
and Maintenance

Educational Game
Development

Visualization
Programming

Figure 2. Classification of courses by course direction

2.2.2 Methods of analyzing curriculum relevance

The relevance of a course is determined by its type, purpose, requirements, content, and credit hours.
The correlation between courses can be explored through quantitative analysis of course grades. For
example, analyze the achievement data using multiple linear regression, and obtain the statistical
relationships and laws between courses by constructing a data model. Based on students’ academic
performance, correlation analysis, analysis of variance, and regression analysis methods are used to
determine the correlation between a course and subsequent courses.

The analysis methods of course correlation are mainly:

1) Simple correlation analysis: simple correlation analysis is a method of analyzing the

correlation between two variables. In curriculum analysis, it is mainly used to analyze the
correlation between different courses under the same category.
 Convergence of Artificial Intelligence Algorithms and Education Fuels Educational Reform in Higher 11
Education Programs

2) Typical correlation analysis: Typical correlation analysis is a method of analyzing the

correlation between one set of data ( X1 , X 2 , X 3 , , X m ) and another set of data
(Y1, Y2 , Y3 ,
, Yn ) . In curriculum analysis, it is mainly used to analyze the overall correlation
between different types of courses.

3) Association rule analysis: association rule analysis is used to find the valuable correlation in
a large amount of data and get the association rules describing the relationship between affairs
so that we can infer the information of another thing from the information of a certain thing.
The degree of interconnection between courses is the main focus of curriculum analysis.

This paper mainly focuses on clustering association rules.

2.2.3 Simple correlation analysis of courses

When there is a quantitative relationship between the things under study, but not like a functional
relationship with definite, stable, and one-to-one corresponding values, this relationship is called
correlation. Correlation is mainly used to describe the relationship between the variables that the
functional relationship can not express, but there is a dependent relationship; it is the use of
appropriate statistical indicators to indicate the strength and direction of the correlation between the
variables and variables.

Correlation is mainly used to study the correlation between things, but it cannot directly reveal the
internal causal relationship. Therefore, if you want to judge whether the correlations have a causal
relationship at the same time, you have to further analyze according to the existing knowledge and
experience.

The correlation coefficient is generally expressed as r which reflects the direction of change and the
degree of closeness of the correlation between things. Its value range is 0 | r | 1 , where the sign of
r indicates the direction of change between the variables, “+” sign indicates that the trend of change
between the variables is the same, with the same increase or decrease, that is, positive correlation, “-”
sign indicates that the direction of change is opposite, that is, negative correlation. The absolute value
of r indicates the closeness of the connection between things.

According to previous research, according to the size of the correlation coefficient can be obtained in
several different degrees of correlation: | r | 0.8 , the two variables are highly correlated,
0.5 | r | 0.8 , the two variables are significantly correlated, 0.3  r | 0.5 , the two variables are
lowly correlated, | r | 0.3 , and the two variables are not correlated. In practical problems, in addition
to calculating the degree of correlation, it is necessary to complete the test of significance, thus
reducing the risk of randomization of the sample data.

Simple correlation analysis is a method of analyzing the correlation between two variables in order
to reveal the degree and direction of the link between the variables. Different measures are used for
various types of variables, three of which are commonly used: Pearson product distance correlation,
Spearman rank correlation, and Kendall’stau-b rank correlation.

Pearson product-distance correlation is mainly used to calculate the correlation between continuous
variables, which requires that the overall variable is normally distributed or can be approximated as
conforming to a normal distribution, and the sample size should be not less than 30 for the Pearson
correlation coefficient to be meaningful.
 12 Jingjing Nie. Applied Mathematics and Nonlinear Sciences, 9(1) (2024) 1-21

1) Pearson product distance correlation coefficient formula:

 ( a − a ) (b − b )
n

i i
rAn = i =1
(16)
 ( bi − b )
n n

 ( ai − a )
2 2

i =1 i =1

In the formula, variable A = ( a1 , a2 , a3 ,, an ) , variable B = ( b1 , b2 , b3 ,, bn ) , a and b are the mean
values of variables A and B respectively, ai and bi are the observed values of variables A and B
respectively, and n is the sample size.

2) Test statistic formula:

rAB n − 2
t= (17)
1 − rAB
2

When | t | tn /2 (or probability p < specified  value), it means that the correlation coefficient r
derived from the sample data is significant and there is a significant correlation between the two
aggregates. When | t | tn /2 (or probability p > specified  value), it means that the correlation
coefficient r derived from the sample data is not statistically significant, and then the two overall
correlations are not significant. It is generally accepted that at  = 0.05 , variables A and B are
significantly correlated, and at  = 0.01 , variables A and B are highly correlated considerably.

Spearman’s rank correlation is mainly used to measure the correlation between non-normally
distributed or ordinal variables such as 1-satisfactory, 2-dissatisfactory, 3-very dissatisfactory, etc.

Kendall’s tau-b rank correlation is mainly a non-parametric method based on ordinal data calculation
by calculating the number of homoscedastic pairs and heteroscedastic pairs and then obtaining the
correlation coefficients, and then testing for significance.

In the analysis of curriculum correlation, each course can be regarded as a variable, so the magnitude
r calculated by the formula can be used to initially determine how closely the courses are linked.

2.2.4 Typical correlation analysis of courses

Typical correlation analysis is a multivariate statistical method for analyzing the overall
connectedness of two sets of things. In general, for one set of variables X1 , X 2 , X 3 , , X m (m  1) and
another set of variables Y1 , Y2 , Y3 , , Yn (n  1) , the combined variables of these two sets of variables
are U and V , i.e., a linear combination of X and Y :

U = a1 X 1 + a2 X 2 + a3 X 3 + + am X m = a X
(18)
V = b1Y1 + b2Y2 + b3Y3 + + bnYn = bY

The algorithms for typical correlation coefficients:

 Convergence of Artificial Intelligence Algorithms and Education Fuels Educational Reform in Higher 13
Education Programs

Cov(U ,V ) Cov ( a X , bY )

 (U ,V ) =  ( a X , bY ) = = (19)
var(U ) var(V ) var ( a X ) var ( bY )

In this study, a correlation rule analysis of the curriculum will be conducted to explore noteworthy
correlations between employer business feedback and student achievement data while in school.

3 Results and Discussion

3.1 Data acquisition and pre-processing

3.1.1 Data acquisition

Using a college design program as a data source, all curriculum system data for each program is
maintained in the faculty management system. The amount of this data is also updated and increased
each year, and the updates are generally as follows:

1) Statistics on the number of students enrolled, the number of students enrolled, and the number
of students changing majors for each new academic year.

2) Statistics on the entry and change of course information for a major in a new academic year.

3) Employers in various industries are changing their requirements for employees. They put
forward new requirements to the schools that cultivate professional competence, and the
school’s professional curriculum needs to be set up in line with the development of society as
well as the requirements of employers. Academic administrators are tasked with revising and
re-entering relevant professional course information and course content that have changed in
the new academic year and semester.

4) The changes mentioned above are not only for new students but also for current students
whose curriculum data will be modified according to the requirements of society and
employers.

5) Other related information of the specialized courses. For example, the education authority of
the university has new requirements for the training program and curriculum, in which case
the data needs to be updated.

In this paper, the data on curriculum, teaching quality, graduates’ abilities, and employers’ evaluation
of a major in a university are selected for analysis. The curriculum framework diagram of this major
is shown in Figure 3. According to the data in the teaching management system, the professional
courses are divided into four categories: industry general courses, professional core courses, job-
specific courses, and job expansion courses.

The data on the setting situation of professional courses, the mastery of students’ professional ability,
and the demand situation of enterprises for the profession are taken as the research objects to carry
out hierarchical association rule data mining. The specific data is as follows:

The first thing to do is to get professional course information. Including course status information
[including the situation of course settings in the talent training program (i.e., the original information
of the course), course changes, the nature of the course, the situation of course hours (recorded in
terms of the total number of semester hours), and the category to which the course belongs.
 14 Jingjing Nie. Applied Mathematics and Nonlinear Sciences, 9(1) (2024) 1-21

Second, feedback information from the enterprise. Including the common software of the enterprise,
the requirement for knowledge mastery, the requirement for professional knowledge, and the
enterprise’s opinion on the improvement of the course.

Professional Courses

Industry General Courses

Job Expansion Courses
Professional Foundation Professional Core Courses Job-specific Courses
Elective Courses
Courses

Figure 3. Framework diagram of the course

3.1.2 Data pre-processing

Handling missing values: The reasons for missing data in the curriculum are mainly due to changes
in the curriculum. Separate handling should be done for different grades, with the exception of the
change in grade courses and deletion of data before the change. For example, the “Indesign” course
is deleted from the Talent Cultivation Program in the academic year of 2020. The class hours of the
course are merged into the “Book Binding Design” course, and the class hours of the “Book Binding
Design” course are changed from the original class hours to the “Book Binding Design” course. For
example, the “Indesign” course is deleted from the talent training program. Its hours were
incorporated into the “Bookbinding Design” Co. The hours of the “Bookbinding Design” course were
changed from 60 hours per semester to 90 hours per semester, so only 90 hours of “Bookbinding
Design” will be kept in the 2020 academic year curriculum system.

Delete Isolated Points: Delete information in the dataset that is out of the ordinary, e.g., A course may
be entered into the course information and found to belong to two different course categories. It is
possible to use the current category, e.g., Photoshop was in the Professional Foundations category
prior to the 2021 academic year but is now classified as a General Industry course.

The results of the preprocessing of data include the preprocessing of specialized courses and the
preprocessing of business feedback. Programs kept only information that was in line with current
college standards for course category duplications.

The processing of feedback from professional courses and companies is as follows:

Course status is categorized into two grades: A “exists” and B “does not exist.”

The nature courses are divided into three levels: “purely practical courses,” B “courses combining
theory and practice,” and “purely theoretical courses.”

Courses are classified into four categories: A “Industry General Courses,” B “Specialized Core
Courses,” C “Job Specific Courses,” and D “Job Expansion Courses.” Courses are classified into four
grades, and the number of hours is based on the actual number of hours.

The common software used by enterprises (the same as the courses set by the major) is divided into
five levels: A “4”, B “3”, C “2”, D “1”, and E “0”.
 Convergence of Artificial Intelligence Algorithms and Education Fuels Educational Reform in Higher 15
Education Programs

The mastery of knowledge is divided into three levels: A “proficient,” B “has a certain foundation,”
and C “cannot use.”

The requirements for professional knowledge are divided into three levels: A “proficient,” B “has a
certain foundation,” and C “not in the requirements.”

The company’s suggestions for improving the curriculum are divided into three levels: A
“Strengthening Theory,” B “Strengthening Practice,” and C “Enhancing Quality.”

After data preprocessing, the data results were obtained. Table 1 shows an excerpt from the course
information sheet, which shows the grade level, course status, nature, and category to which each
course belongs, as well as the class hours. The feedback information table of the enterprise is shown
in Table 2, which determines opinions on improving the mastery level of professional software used
for work and knowledge requirements. Table 3 shows the attribute comparison table, where attributes
such as course status, class hours, knowledge requirements, nature of the course, commonly used
software, improvement opinions, and mastery level are represented by different letter combinations
of abbreviations.

Table 1. Course Information

Course Class
Course title Grade Nature of the course Category
Status hour
Font design 2008 Exist Combining theory with practice Professional core course 90
Font design 2011 Exist Combining theory with practice Industry general courses 60
Indesign 2006 Exist Combining theory with practice Industry general courses 90
Indesign 2012 Exist Combining theory with practice Job development course 90
Bookbinding design 2004 Exist Combining theory with practice Professional core course 60
Bookbinding design 2009 Exist Combining theory with practice Job development course 90

Table 2. Enterprise feedback information

Enterprise Common Knowledge
Mastery degree Improvement suggestions
serial number software requirements
134 4 Good command Basic Skill Strengthen theoretical ability
109 3 Good command Basic Skill Enhance skills
147 2 Basic Skill Basic Skill Enhance skills
 16 Jingjing Nie. Applied Mathematics and Nonlinear Sciences, 9(1) (2024) 1-21

Table 3. Attribute correspondence

Letter Letter Letter
Attribute name Attribute name Attribute name
abbreviation abbreviation abbreviation
Course Status ZT Nature of the course XZ Category LB
Class hour KS Common software RJ Mastery degree CD
Improvement
Knowledge requirements YQ YJ
suggestions

3.2 Validation Analysis of Higher Education Curriculum

The data source file was obtained after the previous data preprocessing steps. The content of the
curriculum data source file section is shown in Table 4. It can be seen that the data includes course
name, grade level, course status, course nature, course category, class time, mastery level,
improvement opinions, and so on. The processed data has basically met the analysis conditions.

Table 4. Part of the data source file

Course Class Mastery Improvement
Course title Grade Nature of the course Category
Status hour degree suggestions
Industry
Theatre and XZ=Combining CD=No YJ=No
2008 ZT=Exist general KS=32
Culture theory with practice requirement requirement
courses
Industry
Drama XZ=Combining CD=No YJ=No
2010 ZT=Exist general KS=32
appreciation theory with practice requirement requirement
courses
Industry
Theatre and ZT=No XZ=Combining CD=No YJ=No
2011 general KS=32
Culture Exist theory with practice requirement requirement
courses
Industry
Drama ZT=No XZ=Combining CD=No YJ=No
2012 general KS=32
appreciation Exist theory with practice requirement requirement
courses
Fundamentals of YJ=Strengthen
XZ=Combining Professional CD=Good
mechanical 2014 ZT=Exist KS=72 theoretical
theory with practice core course command
dynamics ability
Fundamentals of YJ=Strengthen
XZ=Combining Professional CD=Good
mechanical 2016 ZT=Exist KS=72 theoretical
theory with practice core course command
dynamics ability
Basic Electrical YJ=Strengthen
XZ=Combining Professional CD=Good
and Electronic 2014 ZT=Exist KS=72 theoretical
theory with practice core course command
Technology ability
Basic Electrical YJ=Strengthen
XZ=Combining Professional CD=Good
and Electronic 2016 ZT=Exist KS=72 theoretical
theory with practice core course command
Technology ability
Basic Electrical Post-specific YJ=Strengthen
XZ=Combining CD=Good
and Electronic 2014 ZT=Exist direction KS=72 theoretical
theory with practice command
Technology technology ability
Basic Electrical Post-specific YJ=Strengthen
XZ=Combining CD=Good
and Electronic 2016 ZT=Exist direction KS=72 theoretical
theory with practice command
Technology technology ability
… … … … … … … …
XZ=Combining Professional CD=Good YJ=Enhance
AutoCAD 2016 ZT=Exist KS=72
theory with practice core course command skills
Controlling Post-specific
XZ=Combining CD=Basic YJ=Enhance
Technique for 2016 ZT=Exist direction KS=72
theory with practice Skill skills
Programming technology
… … … … … … … …
 Convergence of Artificial Intelligence Algorithms and Education Fuels Educational Reform in Higher 17
Education Programs

The association rule clustering algorithm is used to carry out mining against the source file data using
MATLAB as the software development platform. L is obtained when the minimum support minsup
is set to 0.5, and the value of the minimum confidence minconf is set to 0.4. The frequent 1-term set
is displayed in Table 5. Continue with the subsequent computational step.

Table 5. Frequent 1-itemsets

[Basic Electrical and XZ=Combining CD=Good YJ=Strengthen
ZT=Exist KS=72
Electronic Technology] theory with practice command theoretical ability
[Fundamentals of XZ=Combining CD=Good YJ=Strengthen
ZT=Exist KS=72
mechanical dynamics] theory with practice command theoretical ability
ZT= No
[Theatre and Culture]
exist
[Pro/E application XZ=Combining CD=Basic YJ=Enhance
ZT=Exist
technology] theory with practice Skill skills

Initialize the transaction set C1* as the transaction database, that is, C1*=K. According to the frequent
m-1 item set Lm-1, the frequent m item set Lm is obtained. Generated frequent 2-item sets and support.
Table 6 shows the frequency 2-item set and its support.

From the data in Table 6 Frequent 2-item set and support, it can be seen that “[Fundamentals of
Mechanics], ZT=Existence, XZ=Theory and Practice, CD=Proficiency, KS=, YJ=Enhancement of
Theory” = 0.78992, which means that for the Fundamentals of Mechanics, the course has always
existed in the teaching and learning process at all levels. The nature of the course is a combination of
theory and practice. Employers’ feedback also reflected that for the software students should be
proficient in the application. The nature of the course is a combination of theory and practice, and the
feedback from employers also reflects that students should be proficient in the software application.
Also, it suggests the idea of “enhanced theory.” The frequent 2-item set has a support of 0.78992,
which is greater than the minimum support of 0.5, so the data item will be included in the next
calculation cycle. Frequent generation occurs for the 3-item set and support.

Table 6. Frequent 2-itemsets

Data items list Sustain degree
[Basic Electrical and Electronic Technology], ZT=Exist, XZ=Combining theory with practice,
0.82843
CD=Good command, KS=72, YJ=Strengthen theoretical ability
[Fundamentals of mechanical dynamics], ZT=Exist, XZ=Combining theory with practice,
0.78992
CD=Good command, KS=72, YJ=Strengthen theoretical ability
[Pro/E application technology], ZT=Exist, XZ=Combining theory with practice, CD=Basic Skill,
0.79372
YJ=Enhance skills
…

The data items proceed to the next computation cycle and produce frequent 3-item sets and support,
as shown in Table 7. It can be seen that “[Machine Tool Electrical Control Technology], ZT=Existence,
XZ=Theory and Practice, CD=Proficiency, KS=, YJ=Enhancement of Practical Work” = 0.78846,
which indicates that for the “Machine Tool Electrical Control Technology” course in the teaching and
learning process in all grades has always been the existence of the nature of the course for the
combination of theory and practice, employers feedback in demand to enhance the course must be
proficient, but also proposed to improve the practical work of the course. The nature of the course is
a combination of theory and practice, and the feedback from the employer’s request to enhance the
course must be proficient but also put forward the opinion of strengthening the practical work. The
support for the frequent 3-item set formed is 0.78846, which is greater than the minimum support for
0.5 sets.
 18 Jingjing Nie. Applied Mathematics and Nonlinear Sciences, 9(1) (2024) 1-21

Table 7. Frequent 3-itemsets

Data items list Sustain degree
[Electrical control of machinery tools], ZT=Exist, XZ=Combining theory with practice, CD=Good
0.78846
command, KS=72, YJ=Enhance skills
[Circuit CAD(Protel software)], ZT=Exist, XZ=Combining theory with practice, CD=Good
0.69438
command, KS=72, YJ=Enhance skills
…

In curriculum data analysis, the result to be obtained is the correlation relationship between the
curriculum and each factor, so the most important element is the curriculum, i.e., Obtaining frequent
association rules. The frequent association rules obtained by mining using this paper’s algorithm are
shown in Table 8. Such as association rules [CAXA application technology and practice,
ZT=Exist]=[[XZ=Combining theory with practice, CD=Good command, KS=72, YQ=Basic Skill,
YJ= Enhance skills]], which indicates that the existence of the “CAXA application technology and
practice” course is because of the needs of enterprises, but the quality of teaching, the mastery of
students’ skills and the teaching objectives have a large gap, do not meet the expectations of
enterprises. Therefore, we should not only replan the curriculum and arrange it but also change the
teaching method, combine theory and practice, and increase off-campus practice for this course.

Table 8. Frequent Co-occurrence Rules

[Electrical control of machinery tools, ZT=Exist] = [[XZ=Combining theory with practice, CD=Good command,
KS=72, YJ=Enhance skills]]
[Pro/E application technology, ZT=Exist]=[[XZ=Combining theory with practice, CD=Basic Skill, KS=72,
YJ=Enhance skills]]
[Circuit CAD(Protel software), ZT=Exist]=[[XZ=Combining theory with practice, CD=Good command, KS=72,
YJ=Enhance skills]]
[UG Application technology and practice -1, ZT=Exist]=[[XZ=Combining theory with practice, CD=Good
command, KS=72, YQ=Basic Skill, YJ=Enhance skills]]
[UG Application technology and practice -2, ZT=Exist]=[[XZ=Combining theory with practice, CD=Good
command, KS=72, YQ=Basic Skill, YJ=Enhance skills]]
[Mold CAD/CAM and practical application, ZT=Exist]=[[XZ=Combining theory with practice, CD=Good
command, KS=72, YJ=Enhance skills]]
[CAXA application technology and practice, ZT=Exist]=[[XZ=Combining theory with practice, CD=Good
command, KS=72, YQ=Basic Skill, YJ=Enhance skills]]
…

3.3 Analysis of the effect of curriculum reform

According to the results of the previous course curriculum validation analysis, the higher education
curriculum reform has been implemented.

In order to verify the effect of student training after the reform, two teachers carry out process
evaluation and result evaluation on the teaching effect of the new curriculum setting, take the average
score of the two teachers’ evaluation, and the score data are shown in Table 9. According to the
evaluation table of students’ performance, it can be seen that among a total of 9 groups of students,
there are 5 groups with greater than 80 points, 1 group with 70-80 segments, and 3 groups with 60-
70 segments, which can be concluded that all the students are able to carry out the study of the
professional core courses and complete the basic business of the skilled position under the guidance
of the curriculum resources and teachers. Among them, the students with more than 80 points have
performed better in both the learning process and the final presentation of work results. According to
the observations of teachers during the process, these students are more solid in basic theories and
have sufficient mastery of industry general courses and professional basic courses. Secondly, they
 Convergence of Artificial Intelligence Algorithms and Education Fuels Educational Reform in Higher 19
Education Programs

are more active in thinking, have innovative abilities and expansive thinking, put forward their new
viewpoints and ideas, and are more willing to explore the unknown in the process of practicing. Their
motivation to learn is evidently stronger, and they are capable of overcoming emotional barriers when
confronting difficult problems, solving problems in multiple ways, and working harmoniously in a
group. Fourth, they are capable of mastering both the knowledge and skills points of the core
curriculum and their work presentation is normal and able to demonstrate advanced functions.
Students with scores below 70 are able to follow up on the progress of the class because the
knowledge base is poor in the early stage, and they spend more time supplementing the knowledge
points and basic theories of the degree in the course and do not have more energy to innovate.

Although there are three groups of students with total scores in the 60-70 range, which is relatively
poor, the total scores of all groups are above 60, indicating that the student’s learning of the core
courses and vocational skills courses is effective. The students are able to complete the basic business
of the corresponding positions in their majors and have the ability to work. After the curriculum
reform using the analytical model, the new curriculum effectively improves the quality of teaching in
higher education and better meets the employers’ expectations of the vocational skills level of
graduates, realizing a win-win situation for colleges and universities, enterprises and students.

Table 9. Student score data

Process The t
Groupin Emotions Knowledge Work Appearance Technical Video Total
and exchange
g and values and skills function of work report and pp score
method shows
Group 1 20.25 11.55 20.25 9.45 5.25 6.05 5.85 7.15 85.8
Group 2 21.15 12.75 21.15 9.45 5.75 6.05 5.85 6.05 88.2
Group 3 18 10.5 19.35 7.65 4.2 4.95 4.05 4.95 73.65
Group 4 19.8 11.55 20.25 9.45 5.75 6.05 4.95 6.05 83.85
Group 5 21.15 12.6 21.15 10.35 6.85 6.05 5.4 6.6 90.15
Group 6 17.1 9 16.65 5.4 4.2 3.85 2.7 3.85 62.75
Group 7 19.35 11.25 18.45 9.45 5.75 4.95 5.4 6.6 81.2
Group 8 16.65 8.4 16.2 6.3 4.75 4.4 3.6 4.95 65.25
Group 9 15.75 8.4 15.75 4.5 4.2 4.4 2.7 4.4 60.1

4 Conclusion

This paper analyzes the students’ professional course curriculum, skill mastery, and employers’
requirements using the association rule clustering algorithm to verify the association relationship in
the higher education curriculum system and draws the following conclusions:

1) In the association rule mining, the minimum support degree is set to 0.5, and the minimum
confidence value is 0.4. The association rule obtained from mining [CAXA application
technology and practice, ZT=Exist]=[[XZ=Combining theory with practice, CD=Good
command, KS=72, YQ=Basic Skill, YJ=Enhance skills]] shows that the existence of “CAXA
application technology and practice” course is because of the enterprise’s business. However,
the teaching quality and students’ skill mastery are significantly different from the teaching
goal, which does not meet the expectations of enterprises. Therefore, it is not only necessary
to re-plan the curriculum and arrange it, but also to change the teaching method by combining
theoretical and practical teaching and increasing off-campus practice for this course.
 20 Jingjing Nie. Applied Mathematics and Nonlinear Sciences, 9(1) (2024) 1-21

2) To test the cultivation effect after the curriculum reform, there are 5 groups of students
scoring more than 80 points, 1 group in the 70-80 segment and 3 groups in the 60-70 segment.
All groups of students scored more than 60 points, indicating that they can complete the
basic business of the corresponding positions in this specialty and can work.

After the curriculum reform using the analytical model, the new curriculum effectively improves
the quality of higher education teaching and better meets employers’ expectations regarding
graduates’ vocational skills level.

Funding:

Hubei Provincial Education Science Planning Project, Research on Generative Teaching Design
Based on OBE Concept and Cognitive Load Theory (2021GB074).

References
[1] Xia, Q., Chiu, T. K. F., Lee, M., Sanusi, I. T., Dai, Y., & Chai, C. S. (2022). A self-determination theory
(sdt) design approach for inclusive and diverse artificial intelligence (ai) education. Computers &
education.
[2] Riekki, J., & Mammela, A. (2021). Research and education towards smart and sustainable world. IEEE
Access.
[3] Jiang, Y., & Li, B. (2021). Exploration on the teaching reform measure for machine learning course
system of artificial intelligence specialty. Scientific Programming.
[4] Li, M., & Su, Y. (2020). Evaluation of online teaching quality of basic education based on artificial
intelligence. International Journal of Emerging Technologies in Learning (iJET), 15(16), 147.
[5] Gao, Z., Wanyama, T., & Singh, I. (2020). Project and practice centered learning: a systematic
methodology and strategy to cultivate future full stack artificial intelligence engineers. The international
journal of engineering education(6), 36.
[6] Lechuga, C. G., & Shayan, D. (2023). Three algorithms for grouping students: a bridge between
personalized tutoring system data and classroom pedagogy. International Journal of Artificial Intelligence
in Education(4), 33.
[7] Xiao, S., Shanthini, A., & Thilak, D. (2021). Instructor performance prediction model using artificial
intelligence for higher education systems. Journal of Interconnection Networks.
[8] Carneiro, P., Liu, K., & Salvanes, K. G. (2023). The supply of skill and endogenous technical change:
evidence from a college expansion reform. Journal of the European Economic Association.
[9] Bosman, L., Madamanchi, A., Bartholomew, S., & Byrd, V. (2022). Repeated use of adaptive comparative
judgment to develop student understanding of artificial intelligence in problem based learning
assignments. The international journal of engineering education.
[10] Claudine, Muraraneza, Ntombifikile, Gloria, Mtshali, & Donatilla, et al. (2017). Issues and challenges of
curriculum reform to competency-based curricula in africa: a meta-synthesis. Nursing & Health Sciences.
[11] Li, Q. (2021). The use of artificial intelligence combined with cloud computing in the design of education
information management platform. International Journal of Emerging Technologies in Learning
(iJET)(05).
[12] Hodgson, D., Goldingay, S., Boddy, J., Nipperess, S., & Watts, L. (2021). Problematising artificial
intelligence in social work education: challenges, issues and possibilities. British Journal of Social
Work(1).
[13] Paek, S., & Kim, N. (2021). Analysis of worldwide research trends on the impact of artificial intelligence
in education. Sustainability, 13.
 Convergence of Artificial Intelligence Algorithms and Education Fuels Educational Reform in Higher 21
Education Programs

[14] Wu, J. S., Chien, T. H., Chien, L. R., & Yang, C. Y. (2021). Using artificial intelligence to predict class
loyalty and plagiarism in students in an online blended programming course during the covid-19 pandemic.
Electronics.
[15] Costa, R. D. D., Souza, G. F. D., Castro, T. B. D., Valentim, R. A. D. M., & Dias, A. D. P. (2020).
Identification of learning styles in distance education through the interaction of the student with a learning
management system. Revista Iberoamericana de Tecnologias del Aprendizaje, PP(99), 1-1.
[16] Bauer, E., Greisel, M., Kuznetsov, I., Berndt, M., Kollar, I., & Dresel, M., et al. (2023). Using natural
language processing to support peer‐feedback in the age of artificial intelligence: a cross‐disciplinary
framework and a research agenda. British Journal of Educational Technology.
[17] Chen, L., Chen, P., & Lin, Z. (2020). Artificial intelligence in education: a review. IEEE Access, 8, 75264-
75278.
[18] Park, Y., & Shin, Y. (2021). Tooee: a novel scratch extension for k-12 big data and artificial intelligence
education using text-based visual blocks. IEEE Access.
[19] Maghsudi, S., Lan, A., Xu, J., & Mihaela, V. D. S. (2021). Personalized education in the artificial
intelligence era: what to expect next. IEEE Signal Processing Magazine(3), 38.
[20] Miyamoto, Y. (2017). On the significance of humboldt’s bildung theory for current curriculum reform—
from the series of studies based on humboldt’s bildung theory by d. benner—. Journal of Educational
Research, 21.
[21] Fu, G., & Clarke, A. (2019). Individual and collective agencies in china’s curriculum reform: a case of
physics teachers. Journal of research in science teaching, 56(1), 45-63.
[22] Lambert, K., Alfrey, L., O’Connor, J., & Penney, D. (2021). Artefacts and influence in curriculum policy
enactment: processes, products and policy work in curriculum reform:. European Physical Education
Review, 27(2), 258-277.