https://doi.org/10.31449/inf.v47i9.5222 Informatica 45 (2021) 157–164 157 
An Algorithm for Data Management of Higher Education Based on 
Fuzzy Set Theory - Association Rule Mining Algorithm 
Youmeng Guan 
Xingtai Open University, Xingtai 054000, China 
E-mail: guayou71546@163.com 
Keywords: fuzzy set, higher education, association rule, data-based management 
Received: September 22, 2023 
Data management can enhance the efficiency of higher education management. When combined with data 
mining and other technologies, it can provide a sound foundation for making management decisions. This 
article combined the Apriori algorithm in association rules with fuzzy theory and optimized the FCM 
algorithm to mine fuzzy association rules for student course grades. The results indicated that the 
improved FCM algorithm demonstrated a more effective clustering effect on Iris, and the outcomes were 
closer to the actual values. Applying this method to fuzzy association rule mining could reveal the 
connections between students' course grades. For instance, when students achieved excellent grades in 
the course of computer application fundamentals, their performance in principles of computer 
composition was also good. Furthermore, if they obtained excellent grades in computer application 
fundamentals and good grades in principles of computer composition, their grades achieved in operating 
systems were also excellent. The experimental results validate the reliability of the fuzzy association rule 
mining algorithm, which allows for the discovery of associations between different courses. Consequently, 
it provides valuable support for education and teaching. 
Povzetek: Članek obravnava izboljšanje upravljanja visokega šolstva z upravljanjem podatkov in 
rudarjenjem. Uporabljen je izboljšan algoritem FCM za odkrivanje povezav med ocenami študentov. 
 
1 Introduction 
Under the influence of technological development, the 
field of education and teaching is increasingly shifting 
towards digitization and data-driven approaches [1]. 
However, as colleges and universities expand and their 
systems continue to operate, the volume of data stored in 
the management systems grows annually, putting 
significant pressure on system performance. As a result, 
data mining techniques have been widely adopted in 
various aspects, such as predicting students' grades and 
evaluating the quality of teaching and learning [2]. 
Currently, research in this field has mainly focused on 
predicting student performance and grades, with relatively 
little analysis of the interrelationships between courses. 
However, course arrangement and design are crucial 
aspects of higher education teaching arrangements. In 
order to understand the reliability of data mining 
technique for course correlation analysis, this paper 
studied the application of association rule mining 
algorithms. To enhance mining efficiency, a fuzzy 
association rule combined with fuzzy set theory was 
designed to uncover correlations between students' grades 
in various courses. The goal of this study is to provide 
theoretical guidance for organizing educational and 
teaching work. The study results confirm that the proposed 
method effectively extracts useful data from education 
management systems, improves data-based management 
efficiency, enhances decision-making and management 
capabilities in colleges and universities, and contributes to 
the overall improvement of education and teaching  
 
quality. It is also beneficial for further application of data 
mining technology in educational work. 
2 Related works 
Current research on data mining techniques in educational 
teaching work is presented in Table 1. 
 
Table 1: A summary table of related works 
Literat
ure 
Method Dataset Result 
Barua
h et al. 
[3] 
The MapReduce 
framework based on 
the proposed fractional 
competitive multi-
verse optimization-
based deep neuro-
fuzzy network 
Performan
ce data of 
students 
The mean 
squared 
error, root 
mean 
squared 
error, and 
mean 
absolute 
error 
are .0.3383, 
0.5817, and 
0.3915, 
respectively
. 
Sando
val et 
al. [4] 
A prediction model 
based on low-cost 
variables and a 
sophisticated 
algorithm 
Performan
ce data of 
students 
They 
improved 
the model 
by up to 
12.28% in 
terms of 
root-mean-
square 

158 Informatica 47 (2023) 157–164 Y. Guan 
error. 
Joshi 
et al. 
[5] 
CatBoost - an 
ensemble machine 
learning model  
Performan
ce data of 
students 
The 
accuracy is 
92.27%. 
Fan et 
al. [6] 
A deep learning 
method of 
recommending 
MOOCs to students 
based on a multi-
attention mechanism 
comprising learning 
records attention, 
word-level review 
attention, sentence-
level review attention 
and course description 
attention 
Real-
world data 
consisting 
of the 
learning 
records of 
6,628 
students 
for 1,789 
courses 
and 
65,155 
reviews. 
MOOC 
platforms 
must fully 
utilize the 
information 
implied in 
course 
reviews to 
extract 
personalize
d learning 
preferences. 
3 Association rule mining algorithm 
3.1 Association rules 
Association rule mining algorithms were initially 
employed in supermarket shopping to analyze customers' 
purchasing patterns, enabling retailers to optimize product 
placement and increase sales [7]. Over time, the 
continuous advancement of association rule mining 
algorithms has led to their widespread adoption in data 
analysis across various domains, including healthcare [8] 
and engineering control [9]. 
Association rules can be written as 𝑋 ⇒ 𝑌 , meaning 
that there is a high probability that a record containing 𝑋 
will include 𝑌 (𝑋 is called the former term, and 𝑌 is called 
the latter term). Taking supermarket shopping behaviors 
an an example, suppose: 
of the customers who purchased a coke, 90% 
purchased potato chips (30% of all customers purchased 
both coke and chips). 
In this example, "coke" is the former term and "potato 
chips" is the latter term, "90%" refers to the confidence 
level of the rule, while "30%" refers to its support level. 
According to this rule, coke and potato chips can be 
displayed in similar positions, thus increasing sales. 
Suppose that in database 𝐷 , the set of all items is: 𝐼 =
{𝐼 1
, 𝐼 2
, ⋯ , 𝐼 𝑛 } , and the set of some items is: 𝑇 =
{𝑡 1
, 𝑡 2
, ⋯ , 𝑡 𝑛 }. If there is A ⊂ I, B ⊂ I, and A ∩ B ≠ ∅, 
then an association rule is obtained: A ⇒ B. The following 
rules exist in the association rule. 
(1) Support level: the proportion of records in 𝐷 
containing both A and B in all records, which is written as: 
𝑠𝑢𝑝𝑝 (𝐴 ⇒ 𝐵 ) =
|{𝑇 :𝐴 ∪𝐵 ⊆𝑇 ,𝑇 ∈𝐷 }|
|𝐷 |
.  (1) 
(2) Confidence level: the proportion of records in 𝐷 
containing both A and B in all records containing 𝐴 only, 
which is written as: 
𝑐𝑜𝑛𝑓 (𝐴 ⇒ 𝐵 ) =
|{𝑇 :𝐴 ∪𝐵 ⊆𝑇 ,𝑇 ∈𝐷 }|
|{𝑇 :𝐴 ⊆𝑇 ,𝑇 ∈𝐷 }|
.  (2) 
(3) Frequent term set: if there is 𝑠𝑢𝑝𝑝 {𝐴 } ≥
𝑚𝑖𝑛𝑠𝑢𝑝𝑝 , where 𝑚𝑖𝑛𝑠𝑢𝑝𝑝 is the minimum support level, 
then {𝐴 } is called the frequent itemset. 
(4) Strong rule: If there is 𝑠𝑢𝑝𝑝 (𝐴 ⇒ 𝐵 ) ≥ 𝑚𝑖𝑛𝑠𝑢𝑝𝑝 
and 𝑐𝑜𝑛𝑓 (𝐴 ⇒ 𝐵 ) ≥ 𝑚𝑖𝑛𝑐𝑜𝑛𝑓 , where 𝑚𝑖𝑛𝑐𝑜𝑛𝑓 is the 
minimum confidence level, then 𝐴 ⇒ 𝐵 is called a strong 
rule. 
3.2 Apriori algorithm 
The Apriori algorithm is a classical association rule 
mining algorithm [10]. For database 𝐷 , the sequence of 
steps involved in the mining process of the Apriori 
algorithm is outlined below. 
(1) Database D is traversed to obtain a one-
dimensional set of candidate items. 
(2) Infrequent subsets in the one-dimensional 
candidate item set are pruned to obtain a one-dimensional 
frequent item set. 
(3) The one-dimensional frequent item set is self-
connected to obtain the two-dimensional candidate item 
set. 
(4) Infrequent subsets in the two-dimensional 
candidate item set are pruned to obtain the two-
dimensional frequent item set. 
(5) The k-dimensional frequent term set is self-
connected to obtain the k+1-dimensional candidate term 
set. 
(6) Infrequent subsets in the k+1-dimensional 
candidate item set are pruned to obtain the k+1-
dimensional frequent item set. 
(7) There is no new set of frequent items, and mining 
ends. 
The mining process of the Apriori algorithm with a 
minimum support threshold set at 2 can be illustrated in 
Figure 1, taking a simple dataset as an example. 
 
Figure 1: The mining process of the Apriori algorithm 
As shown in Figure 1, the Apriori algorithm 
continuously scans and prunes the dataset, eventually 
obtaining a 3-dimensional frequent itemset {B, C, E}, at 
which point the algorithm terminates. However, the main 
drawback of the association rule in the case of the Apriori 
algorithm is that it can only deal with discrete data [11] 
and is easy to have the problem of overly hard boundary 

An Algorithm for Data Management of Higher Education Based… Informatica 47 (2023) 157–164 159 
division in the process of dividing continuous attributes 
into discrete intervals. For example, student grades are 
generally divided as follows. 
0 < x < 60 = failed 
60 ≤ x < 70 = pass 
70 ≤ x < 80 = moderate 
80 ≤ x < 90 = good 
90 ≤ x ≤ 100 = excellent 
According to this division, the difference between the 
students whose grades are 59 and 60 respectively is very 
small; however, they are placed in different sets. In 
addition, in real life, there are many non-numerical 
attributes that make this mining algorithm inapplicable. 
Therefore, to better apply association rules for mining 
higher education data, this paper combines them with 
fuzzy set theory. 
4 Fuzzy association rules 
incorporating fuzzy set theory 
4.1 Fuzzy set theory 
In real life, there are many affairs without obvious 
boundaries, such as such as height or shortness, distance 
or proximity, goodness or badness, etc. Fuzzy set theory 
[12] introduces the concept of membership function and 
uses the notion of intermediate transition to realize the 
fuzzy processing of these affairs, which has made 
remarkable achievements in areas like fuzzy control [13], 
pattern recognition [14], and so on. 
Suppose that A is a mapping from domain X to [0,1], 
written as: 
𝐴 : 𝑋 → [0,1], 
then 𝐴 is called the fuzzy set on 𝑋 . 𝐴 (𝑥 ) is the 
membership degree of the fuzzy set. 
4.2 Fuzzy association rule algorithm 
By combining fuzzy set theory with association rules, 
fuzzy association rules can be obtained [15]. Before 
mining the data, the attributes need to be discretized first, 
and the fuzzy C-mean (FCM) algorithm is used [16]. 
Suppose there is dataset  𝑋 = {𝑥 1
, 𝑥 2
, ⋯ , 𝑥 𝑛 } , 𝑗 =
1,2, ⋯ , 𝑛 , the purpose of the FCM algorithm is to divide 
𝑋 into 𝑐 classes and get the clustering center set 𝑉 =
{𝑣 1
, 𝑣 2
, ⋯ , 𝑣 𝑐 } , 𝑖 = 1,2, ⋯ , 𝑐 . Then, the membership 
degree of the 𝑗 -th data belonging to the 𝑖 -th class is written 
as 𝑢 𝑖𝑗
, 𝑢 𝑖𝑗
∈ [0,1], ∑ 𝑢 𝑖𝑗
𝑐 𝑖 =1
= 1. 
The objective function of the FCM algorithm can be 
written as: 
𝐽 (𝑈 , 𝑉 ) = ∑ ∑ 𝑢 𝑖𝑗
𝑚 𝑛 𝑗 =1
𝑐 𝑖 =1
‖𝑥 𝑗 − 𝑣 𝑖 ‖
2
,  (3) 
where 𝑈 = [𝑢 𝑖𝑗
]
𝑐 ×𝑛 is the membership matrix and 𝑚 
is the fuzzy factor. 
The steps of the FCM algorithm are shown below. 
(1) The clustering center is initialized. 
(2) u
ij
 is calculated: u
ij
=
1
∑ (‖x
j
−v
i
‖/‖x
j
−v
r
‖)
2/m−1
 c
r=1
. 
(3) The clustering center is updated: v
i
=
∑ (u
ij
)
m
x
j
n
j=1
∑ (u
ij
)
m
n
j=1
. 
(4) Objective function J and the size of 𝐽 in the last 
iteration are calculated. If 𝐽 is less than or equal to 
termination condition ε or the specified number of 
iterations is reached, then it turns to (5), otherwise it 
returns to (3). 
(5) Clustering result (V, U) is output. 
After attribute discretization, assume that there is an 
arbitrary set of fuzzy attributes 𝑋 = {𝑥 1
, 𝑥 2
, ⋯ , 𝑥 𝑝 }. The 
fuzzy support level of the 𝑖 -th record in fuzzy database 𝐷 𝑓 
for 𝑋 is 𝐹𝑠𝑢𝑝𝑝 𝑖 (𝑋 ). Then, the fuzzy support level of 𝑋 in 
𝐷 𝑓 is: 
𝐹𝑠𝑢𝑝𝑝 (𝑋 ) =
∑ 𝐹𝑠𝑢𝑝𝑝 𝑖 (𝑋 )
𝑛 𝑖 =1
|𝐷 𝑓 |
.  (4) 
The fuzzy association rule is written as 𝑋 𝑓 ⇒ 𝑌 𝑓 , and 
its support level is written as: 
𝐹𝑠𝑢𝑝𝑝 (𝑋 𝑓 ⇒ 𝑌 𝑓 ) =
∑ 𝐹𝑠𝑢𝑝𝑝 𝑖 (𝑋 𝑓 ∪𝑌 𝑓 )
𝑛 𝑖 =1
|𝐷 𝑓 |
.  (5) 
Confidence level is written as: 
𝐹𝑐𝑜𝑛𝑓 (𝑋 𝑓 ⇒ 𝑌 𝑓 ) =
∑ 𝐹𝑠𝑢 𝑝 𝑝 (𝑋 𝑓 ∪𝑌 𝑓 )
𝑛 𝑖 =1
𝐹𝑠𝑢𝑝𝑝 (𝑋 𝑓 )
.  (6) 
Fuzzy association rules follow the same method to 
mine the data and get fuzzy association rules. The FCM 
algorithm has low complexity, is easy to implement, and 
is the most widely used fuzzy clustering method. 
However, the way of randomly determining the initial 
clustering center in the FCM algorithm may bring some 
negative effects on the results [17], making it difficult to 
guarantee their accuracy and determine whether the 
obtained optimal solution is globally optimal. In order to 
improve this problem and increase the reliability of fuzzy 
association rules, this paper uses density function to 
determine the initial clustering center. The FCM algorithm 
only considers distance measurements in its computation 
process. By incorporating a density function, it becomes 
possible to incorporate a measure of the spatial 
distribution of sample points. In a data, if a point is 
surrounded by many other data points, it means that it has 
a high density or is more likely to be the clustering center 
that reduces the effect of isolated noise point on clustering. 
The density function is: 
𝑀 𝑗 = ∑
1
‖𝑥 𝑖 −𝑥 𝑗 ‖
𝑛 𝑗 =1,𝑗 ≠𝑖 , 1 ≤ 𝑖 ≤ 𝑛 , 1 ≤ 𝑗 ≤ 𝑛 .  (7) 
The larger the value of 𝑀 𝑗 is, the more the data points 
distributed around is, and the larger the density is. The 
denser points are used as the initial clustering centers, 
while the region length is introduced to divide the cluster 
centers in order to avoid these points being in the same 
cluster: 
D =
D
max
c
,  (8) 
where 𝐷 𝑚𝑎𝑥 is the distance between the two farthest 
clustered samples in the dataset and 𝑐 is the number of 
categories. 
The procedure of determining the initial clustering 
centers by the improved FCM is shown below. 
(1) The length of the region in dataset D is calculated, 
marking all sample points as searchable. 
(2) Sample point x
i
 with the highest density is used as 
the initial clustering center. 
(3) All sample points in the area where 𝑥 𝑖 locates are 
marked as unsearchable. 

160 Informatica 47 (2023) 157–164 Y. Guan 
(4) It returns to step (2) for c times of iterations to find 
out the sample point with the highest density and use it as 
the initial clustering center. 
By improving FCM, the influence of randomly 
generated initial cluster centers on clustering results can 
be avoided, thus preventing the selection of some isolated 
noise points as cluster centers and reducing result bias. 
5 Student achievement analysis 
results 
5.1 Experimental setup 
The grades of 11 courses for 1,068 students from the 
Computer Science Department in the class of 2019 were 
exported from the student grade management system of 
Xingtai Open University for fuzzy association rule mining 
to eliminate incomplete values and abnormal values. 
Finally, the data of 1,052 students were retained, and some 
of them are shown in Table 2. 
Table 2: Student course grades (unit: point) 
Code Course name 1 2 3 ...... 1052 
A Computer 
application 
fundamentals 
85 91 87 ...... 65 
B Principles of 
computer 
composition 
87 92 89 ...... 64 
C Operating 
systems 
77 93 86 ...... 61 
D Data structure 76 89 87 ...... 62 
E C programming 85 95 88 ...... 69 
F Discrete number 77 85 81 ...... 63 
G Software 
engineering 
67 77 71 ...... 49 
H Computer 
network 
78 85 81 ...... 55 
I Database 
application 
technology 
67 78 73 ...... 55 
J WEB 
development 
basics 
65 77 74 ...... 56 
K Computer 
network security 
technology 
60 70 65 ...... 48 
 
The fuzzification results of students' course grades 
obtained using trapezoidal membership function are 
illustrated in Figure 1. 
0
1
57 63 68 73 78 83 88 93
Failed=5 Pass=4 Medium=3 Good=2 Excellent=1
 
Figure 1: Fuzzification of students' course grades 
The numbers 1-5 were used to assign codes to the 
students' course grades, corresponding to excellent-
failure, respectively. After coding, the data were all 
written in the form of A1, B2, C3, etc., which facilitated 
subsequent mining of fuzzy association rules. 
5.2 Clustering performance analysis 
To determine the effectiveness of the improved FCM 
algorithm, experiments were conducted on the UCI's Iris 
dataset and Seeds dataset [18]. The UCI dataset is a 
commonly used standard benchmark for machine learning, 
and this article selected two subsets from it to validate the 
clustering performance of the improved FCM algorithm. 
The Iris dataset contains 150 samples. The samples were 
divided into three classes with four attributes, such as 
sepal length as displayed in Table 2. They were clustered 
using the traditional FCM and improved FCM algorithms, 
and the results were compared with the actual clustering 
centers given in the literature [19]. The comparison results 
are presented in Table 3. 
Table 3: Comparison of clustering results for the Iris 
dataset 
 Cluste
ring 
Sepal 
length
/cm 
Sepal 
width/
cm 
Petal 
length
/cm 
Petal 
width/
cm 
Literat
ure 
[19] 
1 5.00 3.42 1.46 0.24 
2 5.93 2.77 4.26 1.32 
3 6.58 2.97 5.55 2.02 
Traditi
onal 
FCM 
algorit
hm 
1 5.37 3.46 1.54 0.33 
2 6.16 3.14 4.56 1.31 
3 7.03 3.16 5.88 1.98 
Improv
ed 
FCM 
algorit
hm 
1 5.01 3.41 1.44 0.25 
2 5.99 2.81 4.31 1.29 
3 7.03 2.94 5.57 1.99 
 
From Table 3, it can be found that the gap between the 
clustering results obtained by the traditional FCM 
algorithm and the actual results was large. For example, in 
cluster 1, the sepal length obtained by the FCM algorithm 
was 5.37 cm, while the actual result was 5.00 cm, with a 
difference of 0.37 cm. In contrast, the clustering results 
obtained by the improved FCM algorithm were much 

An Algorithm for Data Management of Higher Education Based… Informatica 47 (2023) 157–164 161 
closer to the actual results, with a smaller gap. The study 
on the Iris dataset showed that the improved FCM 
algorithm had better clustering results and can be applied 
in fuzzy association rules. 
The Seeds dataset consists of 210 samples, which can 
be divided into three classes and seven attributes. The 
traditional FCM and improved FCM algorithms were used 
to cluster the dataset, and the results were compared with 
the actual clustering centers provided by the dataset. The 
results are presented in Table 4. 
 
Table 4: Comparison of clustering results for the Seeds 
datasets 
 Cl
ust
er 
A
re
a 
Peri
met
er 
Com
pactn
ess 
Le
ng
th 
of 
ke
rn
el 
W
idt
h 
of 
ke
rn
el 
Asy
mme
try 
coef
ficie
nt 
Le
ng
th 
of 
ke
rn
el 
gr
oo
ve 
Act
ual 
valu
e 
1 0.
7
6 
0.7
9 
0.69 0.
73 
0.
77 
0.37 0.
76 
2 0.
1
2 
0.1
8 
0.38 0.
19 
0.
16 
0.50 0.
28 
3 0.
3
8 
0.4
2 
0.67 0.
36 
0.
47 
0.26 0.
32 
Trad
ition
al 
FC
M 
algo
rith
m 
1 0.
7
1 
0.6
9 
0.72 0.
75 
0.
71 
0.39 0.
67 
2 0.
1
6 
0.2
1 
0.31 0.
18 
0.
15 
0.55 0.
31 
3 0.
3
1 
0.4
5 
0.68 0.
37 
0.
51 
0.24 0.
29 
Imp
rove
d 
FC
M 
algo
rith
m 
1 0.
7
5 
0.7
8 
0.68 0.
73 
0.
75 
0.39 0.
77 
2 0.
1
3 
0.1
8 
0.37 0.
21 
0.
17 
0.55 0.
28 
3 0.
4
1 
0.4
2 
0.67 0.
35 
0.
49 
0.27 0.
32 
 
From Table 4, it can be observed that similar to the 
results of the Iris dataset, there was a significant 
discrepancy between the clustering results obtained by 
traditional FCM algorithm and the actual cluster centers. 
Taking Cluster 1 as an example, the traditional FCM 
algorithm yielded a length of kernel groove value of 0.67, 
which differed greatly from the actual value of 0.76. In 
comparison, the improved FCM algorithm achieved better 
alignment with the actual values in terms of clustering 
results. The testing conducted on two datasets 
demonstrated that improved FCM performed better in 
clustering tasks. 
5.3 Fuzzy association rule analysis 
Suppose min-supp = 0.3 and min-conf = 0.7. Fuzzy 
association rule mining was performed on students' course 
grades, and the partial results obtained are shown in Table 
5. 
Table 5: Fuzzy association rules 
Former 
term 
Latter 
term 
Support 
level 
Confidence 
level 
A1 B1 0.35 0.97 
J2 G2 0.36 0.95 
K4 G4 0.32 0.88 
A1, B2 C1 0.31 0.76 
C2, D2 I1 0.33 0.85 
E2, F2 I2 0.37 0.77 
H1, J2 G3 0.32 0.91 
I3, J3 K4 0.31 0.81 
A1, C1 G1 0.33 0.84 
A2, B2 K3 0.32 0.71 
 
According to Table 5, the rules obtained are as follows. 
(1) (Computer application fundamentals = 
excellent) ⇒ (principles of computer composition = 
excellent) 
(2) (WEB development basics = good)⇒ (software 
engineering = good) 
(3) (Computer network security technology = pass)⇒ 
(software engineering = pass) 
(4) (Computer application fundamentals = excellent, 
computer application fundamentals = good)⇒ (operating 
systems = excellent) 
(5) (Operating systems = good, data structures = 
good)⇒ (database application technology = excellent) 
(6) (C programming = good, discrete mathematics = 
good)⇒ (database application technology = good) 
(7) (Computer network = excellent, WEB 
development basics = good)⇒ (software engineering = 
moderate) 
(8) (Database application technology = moderate, 
WEB development fundamentals = moderate) ⇒ 
(computer network security technology = pass) 
(9) (Computer application fundamentals = excellent, 
operating systems = excellent)⇒ (software engineering = 
excellent) 
(10) (Computer application fundamentals = good, 
principles of computer composition = good)⇒ (computer 
network security technology = moderate) 
The above rules were analyzed. Taking rule 1 as an 
example, 35% of the students in the database satisfy this 
rule, i.e., i.e., performing excellently in both computer 
application fundamentals and principles of computer 
composition. Moreover, when a student is proficient in 

162 Informatica 47 (2023) 157–164 Y. Guan 
computer application fundamentals, 97% of the students 
are also proficient in principles of computer composition. 
This shows that there is some commonality between the 
two courses and that students who are able to be proficient 
in computer application fundamentals are also able to be 
proficient in principles of computer composition. 
Taking Rule 4 as an example, 31% of students in the 
database satisfy this rule, i.e., when students have 
excellent performance in computer application 
fundamentals and good performance in principles of 
computer composition, they can achieve excellent 
performance in operating systems. Additionally, 76% of 
students demonstrate excellent performance in operating 
systems when they have excellent performance in 
computer application fundamentals and good performance 
in principles of computer composition. This shows that 
computer application fundamentals and principles of 
computer composition serve as the foundation for 
understanding operating systems. If students have a good 
performance in these two courses, they are likely to 
achieve high grades in their operating systems course. 
Take Rule 8 as an example, 31% of students in the 
database satisfy this rule, i.e., when students' performance 
in database application technology is moderate and their 
performance in WEB development fundamentals is also 
moderate, their performance in computer network security 
technology is qualified. When students have a moderate 
performance in database application technology and WEB 
development fundamentals, 81% of them demonstrate 
qualified performance in computer network security 
technology.  This indicates that struggling with database 
application technology and WEB development 
fundamentals will make it more difficult for students to 
learn the course of computer network security technology. 
Some correlations between courses can be identified 
according to the above rules to provide some guidance for 
the school's subsequent curriculum arrangement. For 
example, the study of computer application fundamentals 
and principles of computer composition should be 
preceded by courses in operating system and computer 
network security technology, in order to lay a good 
foundation. Similarly, courses on operating system and 
data structure should be arranged prior to database 
application technology to help students better understand 
the content of database application technology. 
6 Discussion 
The widespread application of data mining 
technology in educational teaching has provided scientific 
and reliable support for the management and decision-
making of educational teaching work, effectively 
improving the efficiency of educational instruction. 
Considering the limitations of current data mining 
technology in course correlation analysis, this paper 
examined the usability of fuzzy association rule mining in 
higher education student course grade correlation analysis 
based on an association rule mining algorithm and 
proposed a method based on an improved FCM algorithm. 
The improved FCM algorithm introduced in this 
paper performed better in clustering by incorporating 
density function design, as demonstrated through 
experiments on two standard test sets. The obtained 
clustering results were closer to the actual cluster centers 
of the data, indicating the reliability of the improved FCM 
algorithm in cluster analysis. Furthermore, when applied 
to fuzzy association rule mining for student course grades, 
this method identified certain associations between 
courses. For example, the performance in database 
application technology and web development basics will 
affect the grades in computer network security technology. 
Similarly, the performance in computer application 
fundamentals also has a certain impact on the grades in  
principles of computer organization. By analyzing these 
fuzzy association rules, we can understand which courses 
should be arranged as foundational courses earlier in the 
curriculum and which courses should be scheduled after 
completing foundational coursework. This provides 
theoretical support for course planners and can also be 
applied to student course selection systems to help 
students choose suitable courses for better learning 
outcomes, thereby avoiding a decline in interest and poor 
learning effectiveness caused by difficulty jumps during 
their studies. In general, using the fuzzy association rule 
mining algorithm to analyze the correlation between 
courses can help optimize teaching plans and improve the 
performance of course selection systems. 
Due to the current application of data mining 
techniques mainly in predicting and evaluating students' 
performance, there has been limited research on course 
association rule mining. Consequently, valuable data in 
educational databases have not been fully utilized. This 
study serves as a starting point for applying fuzzy 
association rule mining in higher education data 
management, demonstrating the effectiveness of this 
method in analyzing course correlations. However, there 
is still room for improvement. For example, there are 
relatively few courses designed in the experiment, and a 
limited number of rules have been discovered. 
Additionally, other data mining algorithms that can be 
applied in this field need to be discussed. In future work, 
it is necessary to further improve the efficiency of mining 
and uncover more useful rules for education, providing 
guidance for educational teaching.  
7 Conclusion 
This paper primarily focuses on data-based management 
in higher education. Fuzzy association rule mining 
algorithms were utilized to analyze the association 
between students' course grades. The results indicated that 
the improved FCM algorithm exhibited a better clustering 
effect compared to the traditional FCM algorithm. 
Additionally, the fuzzy association rule mining based on 
the improved FCM algorithm yielded satisfactory 
outcomes. Moreover, the fuzzy association rule mining 
based on the improved FCM algorithm also achieved 
satisfactory results, revealing rules that were more 
meaningful and accurately reflecting the connections 
between courses. These findings provide a solid 
foundation for actual course arrangement. 

An Algorithm for Data Management of Higher Education Based… Informatica 47 (2023) 157–164 163 
References  
[1] Xie B (2020). Construction of Teacher Culture in 
Applied Colleges under the Background of 
Educational Informationization. Microprocessors 
and Microsystems, 2020, pp. 103486. 
https://doi.org/10.1016/j.micpro.2020.103486 
[2] Ma Y L, Cui C, Nie X, Yang G, Shaheed K, Yin Y 
(2019). Pre-course student performance prediction 
with multi-instance multi-label learnin. Science 
China (Information Sciences), 62, pp. 200-205. 
https://doi.org/10.1007/s11432-017-9371-y 
[3] Baruah A J, Baruah S (2021). Data Augmentation 
and Deep Neuro-fuzzy Network for Student 
Performance Prediction with MapReduce 
Framework. International Journal of Automation 
and Computing, 18, pp. 981-992. 
https://doi.org/10.1007/s11633-021-1312-1 
[4] Sandoval A, Gonzalez C, Alarcon R, Pichara K, 
Montenegro M (2018). Centralized student 
performance prediction in large courses based on 
low-cost variables in an institutional context. The 
Internet and Higher Education, 37, pp. 76-89. 
https://doi.org/10.1016/j.iheduc.2018.02.002 
[5] Joshi A, Saggar P, Jain R, Sharma M, Gupta D, 
Khanna A (2021). CatBoost - An Ensemble Machine 
Learning Model for Prediction and Classification of 
Student Academic Performance. Advances in Data 
Science and Adaptive Analysis: Theory and 
Applications, 13, pp. 1-28. 
https://doi.org/10.1142/S2424922X21410023 
[6] Fan J, Jiang Y, Liu Y, Zhou Y (2022). Interpretable 
MOOC recommendation: a multi-attention network 
for personalized learning behavior analysis. Internet 
Research: Electronic Networking Applications and 
Policy, 32, pp. 588-605. 
https://doi.org/10.1108/INTR-08-2020-0477 
[7] Prahartiwi L I, Dari W (2019). Algoritma Apriori 
untuk Pencarian Frequent itemset dalam Association 
Rule Mining. PIKSEL Penelitian Ilmu Komputer 
Sistem Embedded and Logic, 7, pp. 143-152. 
https://doi.org/10.33558/piksel.v7i2.1817 
[8] Diamond B J, Happawana K A (2022). Association 
rule learning in neuropsychological data analysis for 
Alzheimer's disease. Journal of Neuropsychology, 
16, pp. 116-130. 
[9] Fu L, Wang X, Zhao H, Zhao H, Li M (2022). 
Interactions among safety risks in metro deep 
foundation pit projects: An association rule mining-
based modeling framework. Reliability Engineering 
& System Safety, 221, pp. 1-16.  
https://doi.org/10.1016/j.ress.2022.108381 
[10] Liu X, Sang X, Chang J, Zheng Y, Han Y (2021). 
The water supply association analysis method in 
Shenzhen based on kmeans clustering discretization 
and apriori algorithm. PLoS ONE, 16, pp. 1-21. 
https://doi.org/10.1371/journal.pone.0255684 
[11] Erişti B, Yildirim O, Eristi H, Demir Y (2018). A 
New Embedded Power Quality Event Classification 
System Based on The Wavelet Transform. 
International Transactions on Electrical Energy 
Systems, 28, pp. e2597. 
https://doi.org/10.1002/etep.2597 
[12] Zaki M J, Parthasarathy S, Li W, Ogihara M (1997). 
Evaluation of sampling for data mining of 
association rules. Proceedings Seventh International 
Workshop on Research Issues in Data Engineering, 
pp. 42-50. 
[13] Zoghby H M E, Ramadan H S (2022). Enhanced 
dynamic performance of steam turbine driving 
synchronous generator emulator via adaptive fuzzy 
control. Computers & Electrical Engineering, 97, pp. 
107666-. 
https://doi.org/10.1016/j.compeleceng.2021.107666 
[14] Naranjo R, Santos M (2019). A fuzzy decision 
system for money investment in stock markets based 
on fuzzy candlesticks pattern recognition. Expert 
Systems with Applications, 133, pp. 34-48. 
https://doi.org/10.1016/j.eswa.2019.05.012 
[15] Yavari A, Rajabzadeh A, Abdali-Mohammadi F 
(2021). Profile-Based Assessment of Diseases 
Affective Factors Using Fuzzy Association Rule 
Mining Approach: A Case Study in Heart Diseases. 
Journal of Biomedical Informatics, 116, pp. 103695. 
https://doi.org/10.1016/j.jbi.2021.103695 
[16] Zare M, Koch M (2018). Groundwater level 
fluctuations simulation and prediction by ANFIS- 
and hybrid Wavelet-ANFIS/Fuzzy C-Means (FCM) 
clustering models: Application to the Miandarband 
plain. Journal of Hydro-environment Research, 18, 
pp. 63-76. 
https://doi.org/10.1016/j.jher.2017.11.004 
[17] Siringoringo R, Jamaluddin J (2019). Initializing the 
Fuzzy C-Means Cluster Center With Particle Swarm 
Optimization for Sentiment Clustering. IOP 
Publishing Ltd, pp. 1-6. 
[18] https://archive.ics.uci.edu/ml/datasets/Iris 
[19] Bezdek J C, Hathaway R J, Sabin M J, Tucker W T 
(1987). Convergence theory for fuzzy c-means: 
Counterexamples and repairs. IEEE Transactions on 
Systems Man & Cybernetics, 17, pp. 873-877. 
https://doi.org/10.1109/TSMC.1987.6499296 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

164 Informatica 47 (2023) 157–164 Y. Guan