https://doi.org/10.31449/inf.v48i8.5830                                           Informatica 48 (2024) 111–124  111 
Personalized Learning Resource Push Evaluation Considering Image 
Big Data Processing Technology  
Dafei Wu 
School of Information Engineering, Hunan University of Science and Engineering, Yongzhou, Hunan, 425199, China 
E-mail: yzwdf@126.com 
Keywords: Personalized Learning Resources; Image Big Data Processing; Resource Push  
 
Recieved: March 4, 2024  
With the continuous development of the social economy, the ways of obtaining learning materials and 
learning resources have gradually increased, especially the development of the Internet has enabled 
more high-quality resources to be shared, but it should be noted that it is precisely because of the 
excessive resources, it is relatively complicated to find suitable and interesting learning resources. In 
response to these needs and deficiencies, this paper introduce image big data processing technology, 
characterizes specific learning characteristics using the time decay function by sorting out the business 
logic of personalized learning resource push. It also imports specific learning cognitive levels, and 
matches with the corresponding learning resources, maximizes the quality service of personalized 
Learning facilities, and improves the efficiency of learning resources. The study findings demonstrate 
that the image big data processing technique is effective and can support the push evaluation of 
personalized learning resources.   
Povzetek: Prispevek ocenjuje personalizirano posredovanje učnih virov ob upoštevanju tehnologije 
obdelave slikovnih velikih podatkov. Z uporabo funkcije časovnega razpada za karakterizacijo učnih 
značilnosti in uvozom specifičnih učnih kognitivnih ravni se izboljša učinkovitost in kakovost 
posredovanja učnih virov. Študija potrjuje učinkovitost te tehnologije pri podpori ocenjevanja 
personaliziranih učnih virov.
 
1  Introduction  
In the continual expansion of the economic system, 
individuals are more focused on learning, and the general 
teaching quality and the number of audiences have also 
increased accordingly [1–2]. In addition to the traditional 
paper learning materials, the advancement of online 
technology has contributed to greater online learning 
resources being shared online. Consequently, the Internet 
provides access to corresponding learning resources. On the 
one hand, users can mine the learning resources they are 
interested in, such as MOOCs, Learning Power, etc.; on the 
other hand, users can upload their materials for sharing, so 
that the learning resources form a closed loop. This shows 
that for Internet learning resources, users are both users and 
sharers of learning resources. Starting from individual 
nodes, a fixed connection between the teaching and analysis 
of network learning can be achieved [3–4]. Scholars within 
the industry have carried out in-depth research on this and 
believe that the basis for sharing learning resources should 
be a learning and teaching environment for different users to  
 
achieve personalised applications. Through specific 
teaching methods such as reorganisation and extraction, 
personalised resource learning is realised to maximise the 
exchange and sharing of knowledge and information, to 
achieve personalised learning and sharing, to build a 
corresponding learning environment, and to maximise the 
utilisation of learning resources, aiming to solve the problem 
of limitations in Internet online resource learning [5–6]. 
However, it should be noted that, among the numerous 
learning resources, how to effectively extract the 
information that is of interest and useful to you depends on 
the specific recommendation method. Industry experts have 
carried out extensive research on this topic, frequently 
applying this targeted recommendation method in areas such 
as e-commerce shopping, personalized news browsing, 
personalized book recommendations, and music 
recommendations. For the field of pedagogy, personalized 
specific recommendations can be used to often make 
specific recommendations based on specific recommended 
content, the user's browsing history, and hobbies; or the 
user's learning style and points of interest are used to make 

112   Informatica 48 (2024) 111-124                                                                        D. Wu 
comprehensive and collaborative recommendations. In 
addition, some scholars use the similarity of analogy users 
and collaborative filtering of items to realize the 
recommendation of personalized learning resources; more 
scholars use the effective generation and specific evaluation 
of personalized learning paths, focusing on analyzing the 
knowledge level of users, personal ability and personality 
characteristics for theme-based recommendations. These 
personalised recommendation algorithms achieve the 
purpose and needs of user retrieval to a certain extent, but 
there are still certain limitations. For example, the 
recommendation method mainly focuses on the relationship 
between users and specific resource binary structures, but 
learning resources are not fragmented, independent entities 
but interrelated. In the process of recommendation, it is vital 
to evaluate the link between resources, information systems, 
and users. From the perspective of the learning process, 
learners often overlook the importance of timing. The 
resources used to gain knowledge are constantly 
supplemented, updated, and changed. These temporal 
changes may contain concerns that learners are interested in, 
which may not be in line with the actual learning and 
cognition process. To satisfy the interests of users but also to 
meet the specific cognitive learning resources of users, from 
the beginning to the final mastery, comprehensive 
consideration is required. Improving the use efficiency of 
learning resources, aiming to improve the quality and effect 
of self-learning, and providing users with knowledge that is 
more suitable for their cognitive level are all goals of this 
study. The study accomplishes these goals by conducting a 
business logic analysis of personalized learning resource 
push, representing learning characteristics specifically, 
quantitatively estimating cognitive level, matching existing 
learning resources, and realizing a specific push of online 
learning resources. Table 1 depicts the summary table of 
related works.  
 
Table 1: Related works 
Research Study Methodology Overview Strengths Limitations 
 
 
 
 
 
[27] 
This study delves into the evolving 
intersection of personalized learning 
and Internet technology within the 
education domain. Leveraging 
network technology, the research 
introduces a groundbreaking 
approach: a personalized information 
push system based on 
recommendation.  Algorithms.   
The system excels in 
providing 
personalized learning 
resources, tailoring 
recommendations 
based on users' usage 
history and tag 
attribute 
characteristics. 
While the integration of 
clustering algorithms 
enhances performance, it 
introduced a level of 
algorithmic complexity 
that may require careful 
implementation and 
consideration. 
[28] In response to the rapid evolution of 
big data, this study explored the 
extensive development of artificial 
intelligence teaching systems, which 
have become integral tools for 
independent learning in diverse 
university settings. 
The system offered a 
wealth of learning 
resources and tools, 
contributing to a 
dynamic and enriched 
learning experience. 
The effectiveness of the 
artificial intelligence 
teaching system is 
contingent on 
technological 
infrastructure, potentially 
posing challenges in 
resource-constrained 
environments. 
[29] This study explored the 
transformative potential of artificial 
intelligence technology in advancing 
personalized learning. Leveraging tag, 
index, and intelligent grabbing 
The application of 
virtual reality, 
augmented reality, 
and hybrid reality 
technologies 
Emotion analysis, while 
promising, can face 
challenges in accurately 
interpreting and 
responding to individual 

Personalized Learning Resource Push Evaluation Considering…                          Informatica 48 (2024) 111–124   113 
technology, the research focuses on 
the rapid acquisition of learning 
resources from a complex knowledge 
map. 
contributes to the 
creation of 
personalized online 
learning classrooms, 
fostering immersive 
and engaging 
learning experiences. 
emotions, as emotions can 
be nuanced and 
context-dependent.  
[30] The study recognized the increasing 
demand for personalized learning and 
explored the application of 
Personalized Recommendation (PR) 
technology to enhance learners' 
efficiency. 
The study 
encompasses system 
requirements 
analysis, database 
design, functional 
module design, 
implementation, and 
testing, ensuring a 
comprehensive and 
well-structured 
approach to platform 
development. 
The success of the 
personalized learning 
platform heavily relies on 
the effectiveness and 
sensitivity of the chosen 
CF algorithm, which may 
have limitations in certain 
scenarios. 
[31] In response to the challenges 
presented by the Big-data Era, this 
study focus on the construction of a 
precise personalized learning 
evaluation system to analyze learner 
behaviors and predict personalized 
learning performance. 
The meticulous 
design of first, 
second, and 
third-level indicators 
provides a detailed 
and nuanced 
understanding of 
learners' behavior and 
performance. 
The effectiveness of the 
learning evaluation system 
is contingent on the 
availability and quality of 
learning behavior data, 
which may pose challenges 
in certain contexts. 
 
2  Image big data processing   
The specific types of digital images mainly include the 
following: 
(1) Black-and-white binary image type: Its main feature is 
that the relevant information on the map is either black or 
white, and this contrast is more obvious and stronger. In the 
storage process of the computer, even a single pixel, where 
each pixel contains 1 data bit, can identify at most 2 different 
gray levels. Black-and-white binary image types involve 
managing and analyzing large datasets of monochromatic 
images. This specialized processing deals exclusively with 
binary images, where pixels are either black or white, 
representing distinct visual information. Algorithms and 
techniques for this subset of image processing focus on tasks 
like pattern recognition, edge detection, and shape analysis. 
Within the broader scope of image big data, specific 
processing methods focus on black-and-white binary 
imagery to extract meaningful insights and patterns from 
this distinctive visual data format. 
(2) The type of grayscale: This type of image is often 
divided into multiple gray levels, from black to white, and 
the transition between each gray level is relatively smooth. 
Therefore, this method often results in colors that users do 
not require in this type of image. Generally speaking, each 
pixel includes different long data, and the specific 
composition is realized by different devices. A maximum of 
256 grayscale data sets can be identified. The type of 
grayscale refers to a crucial aspect involving the 
representation of image data. Grayscale images, devoid of 
color, utilize varying shades of gray to depict intensity 
levels. This type of processing involves interpreting and 
manipulating these intensity values, impacting image 
contrast in detail. Understanding and managing grayscale 

114   Informatica 48 (2024) 111-124                                                                        D. Wu 
within the broader context of image big data processing is 
essential for tasks such as image analysis, recognition, and 
enhancement, contributing to more nuanced and 
sophisticated visual data interpretations. 
(3) Colour type: Colour type in big data image processing 
refers to the classification and analysis of images based on 
their colour attributes. It involves extracting, categorizing, 
and processing vast datasets of images to discern colour 
patterns, enabling applications in fields like computer vision 
and image recognition. Understanding color types enhances 
image processing algorithms, aiding in tasks such as object 
detection and segmentation. This subset of image big data 
processing plays a pivotal role in decoding and utilizing the 
rich visual information embedded within diverse datasets. In 
an RGB color image, each pixel can express 16777216 
colors.  
On a specific computer, the representation of an image often 
includes drawing points and short polylines. The method of 
drawing points is to analyze the luminous discrete points on 
the data display screen and effectively combine the pixel 
points on the screen; the method of short polylines can 
realize specific grid image analysis [7–8]. 
For an image, the figure is the distribution response of the 
reflected and projected light of the object; the picture is 
human visual sensory cognition, so the figure exists in 
essence; the picture is the description of the sense; and the 
image realizes the effective integration of the essence and 
the senses. It is not enough to only regard the image as a 
two-dimensional or three-dimensional color distribution; the 
influence of human emotions and psychological factors also 
needs to be considered in specific image processing. Image 
processing is the effective processing and analysis of image 
information to achieve practical applications. 
For computer image processing, it has three main 
characteristics, as follows: 
❖ First, the image processing technology is 
comprehensive. In the corresponding computer 
processing process, basic knowledge of 
mathematics, physics, optics, etc. is often 
involved. At the same time, it also includes 
several technologies such as projection, 
photography, and display. The computer has 
become the main image processing method, so it 
is extremely important. 
❖ The hardware of the computer improves the 
accuracy of the calculation. Depending on the 
actual application, we can increase the accuracy of 
image processing, which in turn improves the 
efficiency of the computer and the corresponding 
calculation accuracy. 
❖ Personalized customization and strong flexibility. 
For image processing, it can effectively process 
any range and any size of the image. 
In the actual personalized learning system, teachers and 
students are important components. Teachers are 
responsible for the effective construction and management 
of corresponding online resources and knowledge bases, and 
they can use ontology to carry out this task. The main 
process is to first construct two parts: the semantic library 
and the resource library. The semantic library is represented 
by the ontology concept to represent the specific knowledge 
points, and the attributes of the ontology are used to connect 
the knowledge points to form a specific semantic layer. The 
corresponding resource library is constructed by learning 
objects, which encapsulate fixed learning resources 
correspondingly, and conducts analysis through unique 
identification and positioning [9–10]. 
When students carry out specific learning, the system 
records according to the student's behavior, uses the learner 
to analyze the changes in the engine, realizes the specific 
learning process and effective access to online resources, 
and constructs the learner's knowledge structure and the 
corresponding cognitive level feature model. The 
corresponding recommendation engine is based on specific 
learning features and labeled learning resources, 
recommends according to specific recommendation 
algorithms, and uses the corresponding sorting order to 
recommend online resources, as shown in Figure 1: 

Personalized Learning Resource Push Evaluation Considering…                          Informatica 48 (2024) 111–124   115 
 
Figure 1: General architecture of a personalized learning 
system. 
2.1 Definition of symbol  
To effectively process big data, first define the initialization 
of symbols, as shown below:  
(1)
i
s : learner i, which means the learner with the unique 
identifier i;  
(2) S: learner set  
12
, , ,
m
S s s s = , represents the set 
composed of m learners;  
(3)
j
o : used to represent the object of learning, that is, used 
to identify a unique object;  
(4) O: the object set of learning, utilize the learning object 
to carry out the constitution of the concrete set;  
(5)
k
c : k used to identify the concept;  
(6) C: Concept set
 
12
, , ,
p
C c c c = , which represents 
the set composed of p concepts.  
2.3 Formal representation of learner feature model  
The formalization of the learner's specific feature model is 
to realize the abstraction of specific learning behavior and 
use the corresponding learning process for analysis. The 
visualization of the feature model is an important part of 
the whole process, and it is the basis of personalized 
recommendations. From the perspective of specific 
recommendations, the specific cognitive structure and level 
of learners can be achieved, mainly focusing on two 
specific perspectives, including the following:  
It represents an effective transmission of learning elements. 
Learning is a dynamic process, during which learners can 
use different periods to learn content, and are constantly 
changing according to their knowledge structure. For 
example, if the learners already have a certain knowledge 
base, they can directly enter the subsequent learning 
without being bound by a specific knowledge point. This 
kind of transfer of interest points that can effectively 
characterize the learner needs to be captured effectively. To 
realize the characterization of the time of interest points, a 
corresponding index can be introduced as a specific 
variable of time decay[11-12].  
Another aspect is the progressive nature of learning. 
According to the standards of human cognition and the 
concept of learning resource setting, the corresponding 
personalized recommended learning resources should 
conform to the specific cognition level of users. If the 
recommended learning resources are too complex or too 
simple, they cannot meet the needs of specific users, nor 
allow the user to truly grasp the specific situation. 
Therefore, a specific knowledge point can be used as an 
evaluation index of a specific ability to measure the user's 
effective grasp of the knowledge point or this type of 
knowledge point. The specific comprehensive 
considerations mainly include two categories: one is the 
correct rate of the user's answer, and the other is the 
difficulty of the question setting. The evaluation results are 
quantified into {1, 2, 3, 4, 5, 6} to achieve specific 
quantitative analysis.   
2.4 Labelling of learning resources  
Annotating semantic information for corresponding 
learning resources can effectively enhance the connectivity 
between learners and resources. The semantic features 
brought by this can be specifically expressed by formula 
(1):  
( ) ( ) ( )
 
, , 1 ,
j y j y y j
V o c dif o c x t c Con o =      
                                        (1) 

116   Informatica 48 (2024) 111-124                                                                        D. Wu 
From the specific architecture, it can be concluded that the 
recommendation engine is the core of the whole process, 
and the recommendation algorithm of online learning 
resources is the specific work content realized by the 
recommendation engine, which is mainly based on the 
association between learners and the corresponding 
learning resources. Meanwhile, it must be based on the 
feature representation of the integrated learner and the 
annotation of the corresponding online learning resources.  
The essence of learning is a specific dynamic change 
process, which is sensitive to time factors, so it is necessary 
to fully consider specific time factors. For example, 
learners need to review the knowledge points they have 
already mastered while avoiding repetitive learning. In the 
real recommendation process, deeper knowledge points or 
more application fields should be recommended; if the 
original knowledge points continue to be pushed, it will be 
counterproductive. Thus, a time decay function is used for 
characterizing the actual variation of learning points of 
interest.  
2.5 Time decay function  
To effectively use the corresponding function to 
specifically characterize the time series characteristics of 
learning resources, the fluctuation of the time series can 
have a specific impact through the function. For the 
function value, the smaller the value, the longer the time, 
and vice versa, the closer the time, the calculated function 
value is. Therefore, the specific calculation of the time 
weight can be specifically expressed by formula (2):  
            (2) 
Among them, the last time point of the learner is denoted 
by ( ) ,
i nearest
tim s c .  
After comparative analysis, it can be seen that when the 
function calculation value of the decay function reaches 1, 
it indicates that the interest of the learning point can best 
express the maximum learning interest point; if the 
function value gradually decreases with the specific time 
change, it reflects the learning activities in the recent period. 
The in recent corresponding learning objects are more 
likely to be recommended, more in line with specific logic, 
and have corresponding rationality.  
2.6 The matching operator between learners' 
cognitive level and learning resource difficulty  
For learning resources, the specific difficulty is related to 
the cognitive level of the learner, and the specific matching 
value is higher, and conversely, the matching value is 
reduced. Therefore, the matching between the interests of 
learners and online learning resources can be specifically 
calculated and expressed by formula (3):  
                        (3) 
From the calculation of formula (3), it can be found that if 
the specific difficulty and cognitive level are equal, the 
corresponding value is the largest; the larger the difference 
between the difficulty value and the cognitive level the 
smaller the corresponding calculated value.  
2.7 Calculation of matching degree between 
learners and learning resources  
The corresponding parameter settings are initialized, by 
calculating the matching degree 
( )
_,
ij
Rec Val s o of 
the learner 
i
s and the learning resource
j
o , as shown in 
formula (4):  
 
            (4) 
In the formula, α and β are adjustment factors, and the 
values of the two must satisfy ; 
In the high-dimensional space, the classification 
hyperplane that separates the two types of samples without 
error satisfies, and the specific calculation is shown in 
formula (5):  
0, ,
n
x b x R b R   + =                 (5) 
Through the normalized analysis of the parameters, the 
specific classification interval analysis is carried out, and 
the specific analysis of the classification samples is carried 

Personalized Learning Resource Push Evaluation Considering…                          Informatica 48 (2024) 111–124   117 
out for the classification line, and the specific calculation is 
shown in formula (6):  
( ) 1 0, 1,2, ,
ii
y x b i n   + −  = 

       (6) 
Based on formula (6), the formula is converted, and the 
specific calculation is as shown in formula (7):  
( ) ( )
2 11
22
     = =                (7) 
Based on formula (7), the constraint conditions are 
calculated, as shown in formula (8):  
1
0, 0, 1, 2, ,
n
i i i
i
y a a i n
=
=  =

           (8) 
Under the condition (4), find the maximum value of the 
following functions for
i
a , and the specific calculation is 
shown in formula (9):  
( ) ( )
1 , 1
1
2
nn
i i j i j
i i j
Q a a a a x x
==
= − 

         (9) 
The optimal classification function obtained by solving is, 
the specific calculation is shown in formula (10):  
( ) ( )   ( )
1
sgn sgn
n
i i i i
i
f x x b a y x x b 
=

=  + =  +



       
                                       (10) 
Based on formula (10), the conditions are relaxed, as 
shown in formula (11):  
( ) 1 , 1,2, ,
i i i
y x b i n   +  − =        (11) 
At the same time, a penalty term is introduced into the 
objective function, and the specific calculation is shown in 
formula (12):  
( )
2
1
1
,
2
n
i
i
C    
=

 = +



           (12) 
Based on formula (12), the setting transformation of 
constraints is carried out, as shown in formula (13):  
1
0,0 , 1,2, ,
n
i i i
i
y a a C i n
=
=   =

      (13) 
The inner product in high-dimensional linear space can be 
specifically calculated by formula (14):  
( ) ( ) ( )
,
i j i j
K x x T x gT x =              (14) 
The three kernel functions that are currently used are as 
follows[13-14]:  
(1) The specific calculation of the polynomial inner product 
function is shown in formula (15):  
( ) ( )
,1
q
i j i j
K x x x gx

=+

              (15) 
(2) The specific calculation of the radial basis function is 
shown in formula (16):  
2
2
, exp
ij
ij
xx
K x x

 
−
 
=−
 
 
          (16) 
(3) The specific calculation of the S-type inner product 
function is shown in formula (17):  
( ) ( ) ( )
, tanh
i j i j
K x x v x gx c =+          (17) 
Personalized learning resources based on students' 
self-learning feedback can better support students' 
self-improvement and self-learning, and promote their 
further development. The main online learning resources 
include two aspects:  
(1) Students independently find and retrieve online learning 
resources that they are interested in;  
(2) According to the corresponding learning evaluation 
information, carry out effective recommendations of 
students' online learning resources, and carry out resource 
allocation processing according to the actual situation, as 
shown in Figure 2:  

118   Informatica 48 (2024) 111-124                                                                        D. Wu 
 
Figure 2: Personalized learning evaluation and resource 
push. 
The specific online learning resource push needs to be 
determined according to the actual learning goals. The 
complete process mainly includes specific steps such as 
data sample collection, data preprocessing, training 
learning, specific evaluation classification, and pushing 
learning resources.  
3  Simulation experiment  
To effectively verify the effectiveness of the image big data 
processing technology, this paper selects the corresponding 
samples for processing and analysis and realizes the push 
analysis of personalized online learning resources.  
3.1 Description of algorithm experiment 
process  
(1) Constructing the knowledge base of "C language 
programming" course  
①Build a semantic library  
When constructing the semantic database, the concepts are 
expressed by the names of knowledge points, and some 
specific knowledge points are shown in Table 2.  
Table 2: Knowledge point collection (part). 
Name of knowledge point  Name of knowledge point  
Type of data  Character string  
Expression  Function  
Input and output functions  Preprocessing command  
Select statement  Pointer  
Loop statement  Structure  
Array  Document  
 
②Annotation Repository  
A specific resource library includes specific learning 
objects. The standard of data collection can be described 
concerning specific metadata, and the online learning 
resources encapsulated by it can be identified and 
effectively located through the network to achieve specific 
semantic labeling [15-16]. Some online learning resources 
are shown in Figure 3. According to a certain period of 
learning, specific learner feature vectors can be obtained.  
0
2
4
6
8
10
formal features
S1
S2
S3
S4
S5
 
Figure 3: Learner formalization features. 
According to the corresponding time function decay factor, 
the specific data is listed as shown in Figure 3. 
It is evident from the findings in Figure 3 that the 
corresponding time function can better reflect the dynamic 
transfer characteristics of learning. According to the 
effective learning time point, the analysis of the 
characteristics is carried out. The greater the timeliness, the 
higher the possibility of being recommended. On the 
contrary, if the calculated value of the time function is 
smaller, the corresponding online resource is less likely to 
be recommended.  
The matching operator of cognitive level and difficulty 
better reflects the cognitive principle of gradual learning. 
For the cognitive matching operator, when the difficulty of 
the learning resource is equal to the cognitive level, its 
value is 1. In this case, the online learning resources are 
more likely to be pushed; on the contrary, if the difference 
between the difficulty value and the corresponding 

Personalized Learning Resource Push Evaluation Considering…                          Informatica 48 (2024) 111–124   119 
cognitive level is large, the corresponding recommendation 
probability is smaller [17-18].  
The learning resources generated by the recommendation 
algorithm conform to the teaching experience of subject 
teachers.  
After specific analysis, in the process of selecting 
corresponding indicators and constructing a specific 
evaluation model, it is still unable to reflect the real 
situation of specific learners intuitively and effectively. 
Secondly, for the coverage of samples, the data samples 
have certain limitations and cannot fully represent all 
situations; finally, the specific evaluation of learning is 
always artificial and subjective, so the evaluation systems 
constructed by different experts are different, unable to 
achieve the objective recognition rate as high as the image 
big data processing technology.  
However, such a model has greatly improved the specific 
degree of automation and effectively improved the 
performance of automatic learning, personalized learning, 
and adaptive learning. At the same time, the complexity of 
the algorithm is relatively low, and it has the performance 
of real-time computing.  
4  Resource acquisition form  
It can be seen from the existing research results that more 
students tend to use professional learning applications 
APPs to search for and learn personalized resources. From 
the specific data samples, it can be seen that students still 
rely too much on traditional teaching classrooms for 
teaching resources, especially English teaching resources. 
Teachers in schools generally publish corresponding course 
resources through information-based teaching methods, 
which has become a common way to obtain online 
resources, specifically as shown in Figure 4:  
0%
20%
40%
60%
80%
100%
proportion
Learning class
using APP
software
browser web
page
WeChat
public
account
Short video
APP
other 
Figure 4: Acquisition form of English mobile learning 
resources for vocational college students. 
4.1 Purpose of usage of resource  
For specific students, the purpose of applying specific 
English learning resources is mainly as shown in Figure 5, 
in which the thirst for new knowledge accounts for the vast 
majority. In addition, the second purpose is to expand 
English knowledge outside the classroom, thus indicating 
that students even more expect to acquire knowledge of 
English outside the curriculum.  
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
70.00%
80.00%
proportion
learn new
knowledge
Contact
consolidation
Expand
English
In-class
learning
Fragmentatio
n promotion
 
Figure 5: The purpose of using English mobile learning 
resources for vocational college students. 
4.2 Motivation of usage of resource  
From the results in Figure 6, it can be seen that the 
student's specific interest and love for English is more that 
learning English is the main driving force for acquiring 
mobile resources, indicating that the design and 
arrangement of English learning tasks will drive students to 
use mobile learning resources.  
0.00%
20.00%
40.00%
60.00%
80.00%
proportion
interested in
English
interested in
culture
For further
studies and
exams
to complete
the task
other
 
Figure 6: Motivation for vocational college students to use 
English mobile learning resources. 

120   Informatica 48 (2024) 111-124                                                                        D. Wu 
4.3 Type of resource requirement  
Elemental learning of resource content is a core component 
of online resources. Using the analysis of students' needs 
for learning English resources, we can clearly understand 
how to integrate, pool, and drive learning resources through 
image big data processing, to match resources and needs 
and improve the utilization rate and effect of English 
learning [19-20]. Specifically, as shown in Figure 7. In 
addition, due to the different foundations of students, it is 
necessary to allow students to carry out effective review 
and development through corresponding resources.  
0.00%
20.00%
40.00%
60.00%
80.00%
proportion
test-taking
resources
Learning
Resources
Professionally
Relevant
English
Resources
Culturally
relevant
English
language
resources
 
Figure 7: Types of English mobile learning resource 
requirements for vocational college students. 
4.4 Resource presentation form  
The image big data processing of the sample data shows 
that the students are more visual learners, and for the online 
learning resources of English, the form data presented is 
also more consistent. More students gravitate to the 
multiple resource formats of video, while audio and 
interactive methods are relatively unpopular (Figure 8).  
0.00%
20.00%
40.00%
60.00%
80.00%
proportion
Graphics
audio
video
Game
animation
interactive test
 
Figure 8: Presentation form preference of vocational 
college students' english mobile learning resources. 
4.4 Personalized resource push research fits 
the development trend of information 
technology   
In the specific development process, how to integrate the 
corresponding technology and teaching, promote the 
comprehensive development of students, solve the key and 
difficult problems in the process of various teaching 
development, motivate the universalization and fairness of 
education, and improve the effective and steady 
improvement of quality of education. By constructing the 
personalized push of corresponding learning resources, the 
personalized and self-organized development of online 
learning resources can be realized, the self-cognition of 
students can be maximized, and the development trend and 
future requirements can be kept up [21-23].   
The abundance of online learning resources limits the 
energy of individual students, making it impossible for them 
to browse, learn, and explore them all. As a result, while 
individualized learning materials might save time and 
money, their availability may be limited. As a result, it is 
required to conduct assessment and analysis based on 
particular indicators, enhance specific indicators for 
evaluation, promote the efficient use of learning resources, 
and encourage the enhancement of college students' learning 
levels [24].   
4.5 System push and customization should be 
adjusted dynamically centring on the learners  
For online learning resources, digital learning resources 
mainly push online resources of interest to specific users 
from the perspective of learners, such as learning interest, 
cognitive level, and specific learning preferences. During 
the actual process of push, it is necessary to take the learner 
as the specific center and make dynamic push adjustments 
according to the feedback information of the learner, to 
achieve the satisfaction of the specific learner. Taken 
together, personalized push needs to include the following 
types of functions [25-26]:  
❖ Analysis of Learners  
The personalized push system actively pushes information 
resources that are of interest to users or that suit their needs 
according to the user's interests, hobbies or professional 
characteristics, and other personalized characteristics.  
❖ Self-update of user interest model  

Personalized Learning Resource Push Evaluation Considering…                          Informatica 48 (2024) 111–124   121 
College students' demand for digital learning resources has 
dynamic characteristics in terms of resource content and 
type.  
❖ Behavior records  
The personalized push system is an implicit score that can 
only be obtained by analyzing the online behavior of 
college students.  
❖ Toolset and Demand Judgment  
The personalized push system should provide a complete 
set of tools to predict the list of digital learning resources 
that college students are interested in based on 
recommendation algorithms.  
4.6 Statistical analysis using ANOV A test  
Analyzing the variations Among the acoustic response of 
the faulty samples and the pristine sample was necessary to 
assess our hypothesis. Analysis of variance (ANOV A) was 
the statistical method. At a confidence interval level, the 
second stage is to expand English knowledge outside the 
classroom, thus indicating that students even more expect 
to acquire knowledge of English outside the curriculum 
specified for ANOV A test input circumstances. An ANOV A 
test analyzes if information from multiple instances of a 
variable shares a similar mean. In essence, the ANOV A test 
can determine if various groups of an independent variable 
have distinct impacts on the reaction variable. ANOV A may 
be formally described as a linear model with multiple input 
sets and i answers per group, as shown in Equation (18).   
               (18) 
Here,    is the response of the 
 
value are group of 
English language the mean group was calculated and Errors 
at random with a mean of zero and fixed values. . The 
ANOVA test total variance of the number was calculated. To 
test for a variance in the group's in equation, divide the 
results into the subsequent two parts:  (19). 
                                                  
                                                 (19) 
Difference group of means from the overall mean,  
 variation among roupings wherein y is the overall 
mean of each group  and z is the overall sample mean. 
The disparity between each group's observations and the 
group mean estimations of variation within a skill process 
group.  Here n is the sample size of jth value where 
i=1,2,…N. An ANOV A test is used to compare the variation 
among the groups of variation within the groups. The ratio 
within the group significantly high outcomes in low values 
of k number of measures test analysis (k − N, N − 1) degrees 
of freedom as presented in Equation (20). 
                      (20) 
The sample tests were conducted to compare the Acoustic 
reaction of the actuation amongst the harmed specimens. 
pristine position, then learning tests were analyzed and 
teachers in schools generally publish corresponding course 
resources through information-based teaching methods.   
5  Discussion  
Artificial intelligence teaching systems are contingent on 
technological infrastructure, potentially posing challenges 
in resource-constrained environments, and the success of the 
personalized learning platform heavily relies on its 
effectiveness and sensitivity. These are the limitations of the 
related works. Comparing our results to the related work 
limitations, our results provide better performance and 
efficiency.  
6  Conclusions 
With the development of the social economy, effective 
access to online learning resources has become increasingly 
important. Given these needs and deficiencies, based on 
image big data processing technology, this paper sorts out 
the business logic of personalized learning resource push, 
uses the time decay function to evaluate the transfer of 
learning interest points, and realizes the matching of 
cognitive level and learning difficulty according to the 
application difficulty. The service of personalized online 
resources is optimized to improve the efficiency of learning 
resources. The simulation experiment results show that the 
image big data processing technology is effective and can 
support the push evaluation of personalized learning 
resources. Limited generalizability due to specific image 
data requirements, potential resource constraints, and 
reliance on advanced technology infrastructure. For future 
research, explore real-world applications, integrate AI 
advancements, and enhance adaptability for dynamic 

122   Informatica 48 (2024) 111-124                                                                        D. Wu 
curriculum changes in the personalized learning resource 
push.  
Data availability 
The data used to support the findings of this study are 
available from the corresponding author upon request. 
 
Conflicts of interest 
The authors declare no conflicts of interest 
Funding statement 
This research was supported by the Hunan University of 
Science and Engineering Application Characteristic 
Discipline Construction Project, Thirteenth Five-Year Plan 
Specialty Comprehensive Reform Pilot Project of General 
Colleges and Universities in Hunan Province ([2016] No. 
276). 
References 
[1] Wan S, Niu Z, 2018, An E-learning Recommendation 
Approach Based on the Self-Organization of Learning 
Resource[J]. Knowledge-Based Systems, 
160(2):56-63. 
[2] Mehta, Clarke, P S, et al, 2016. An assessment of 
student experiences and learning based on a novel 
undergraduate e-learning resource. [J]. British dental 
journal,3(1):10-18. 
[3] Allen L ,  Eagleson R ,  Ribaupierre S, 2015. 
Evaluation of a Virtual 3D Learning Resource in 
Neuroanatomy for Undergraduate Medical 
Students[J]. The FASEB Journal, 29(1):109-117. 
[4] Ring A ,  Mansi J ,  Harari D , et al. 2017, Problem 
Solving in Older Cancer Patients: A case-study based 
reference and learning resource[J]. Anticancer 
Research, 37(3):1543-1550. 
[5] Deformations S O , Ławrynowicz, Julian, 2016. 
System And Method For Efficient Use Of Radio 
Resources For Push-To-Talk Services In Mobile 
Wireless Communications Systems[J]. 
Hydrobiologia, 746(3):1-13. 
[6] Ellis R . 2021, 114Reducing Barriers to Learning 
With eBooks: An Internationally Accessible and 
Affordable Resource[J]. British Journal of Surgery, 
2(2):1-8. 
[7] Wan S ,  Niu Z, 2018,  An E-learning 
Recommendation Approach Based on the 
Self-Organization of Learning Resource[J]. 
Knowledge-Based Systems, 160(1):176-184. 
[8] Yang L ,  Yi Y, 2015, Research for Personalized 
Recommendation of Learning Resource in Mobile 
Computing Context[J]. Lecture Notes in Electrical 
Engineering, 352(2):309-315. 
[9] Zhang L ,  F  Wang,  Qi A, 2017,  Construction 
of Interactive Teaching System for Exercise Training 
Based on Education Video Resource Push 
Technology[J]. International Journal of Emerging 
Technologies in Learning, 12(7):98-107. 
[10] Asheralieva A ,  Niyato D, 2019,  Learning-Based 
Mobile Edge Computing Resource Management to 
Support Public Blockchain Networks[J]. IEEE 
Transactions on Mobile Computing, 2(99):1-10. 
[11] Maura C ,  Assaf P ,  Mussa-Ivaldi F A , 2015, 
Learning to push and learning to move: the adaptive 
control of contact forces[J]. Frontiers in 
Computational Neuroscience, 9(2):10-17. 
[12] Englert P, Vien N A, Toussaint M, 2017, Inverse 
KKT: Learning cost functions of manipulation tasks 
from demonstrations[J]. International Journal of 
Robotics Research, 36(13):1474-1488. 
[13] Yu L ,  Li Z ,  Zhong Y , et al, 2020, When QoE 
meets learning: A distributed traffic-processing 
framework for elastic resource provisioning in 
HetNets[J]. Computer networks, 167(11): 1-13. 
[14] Wan Q ,  Yang X ,  Chen G, 2018,  New 
Scheduling Algorithm for Mobile Teaching Cloud 
Resource Push[J]. International Journal of Emerging 
Technologies in Learning, 13(7):17-26. 
[15] Shafqat S ,  Fayyaz M ,  Khattak H A , et al, 2021 
Leveraging Deep Learning for Designing Healthcare 
Analytics Heuristic for Diagnostics[J]. Neural 
Processing Letters, ,2(1):190-197. 
[16] Zhou Z ,  Yang S ,  Pu L J , et al, 2020 CEFL: 
Online Admission Control, Data Scheduling and 
Accuracy Tuning for Cost-Efficient Federated 
Learning Across Edge Nodes[J]. IEEE Internet of 
Things Journal, , 2(99):1-10. 
[17] Li R ,  Zhao Z ,  Sun Q , et al, 2018, Deep 
Reinforcement Learning for Resource Management in 
Network Slicing[J]. IEEE Access, 3(2):74429-74441. 

Personalized Learning Resource Push Evaluation Considering…                          Informatica 48 (2024) 111–124   123 
[18] Ahmed S ,  Kamal U ,  Hasan M K, 2020. 
DSWE-Net: A deep learning approach for shear wave 
elastography and lesion segmentation using single 
push acoustic radiation force[J]. Ultrasonics, 
110(2):1-8. 
[19] M Kopicki, S Zurek, R Stolkin, 2017,. Learning 
modular and transferable forward models of the 
motions of push manipulated objects[J]. Autonomous 
Robots, 3(2):10-16. 
[20] Yang T ,  Wu D ,  Fang H , et al. 2019,  
Distributed Energy Resource Coordination Over 
Time-Varying Directed Communication Networks[J]. 
IEEE Transactions on Control of Network Systems, 
2(99):1-10. 
[21] Narudin F A ,  Feizollah A ,  Anuar N B , et al. 
2016. Evaluation of machine learning classifiers for 
mobile malware detection[J]. Soft Computing, 
20(1):343-357. 
[22] Ridge B ,  Leonardis A ,  Ude A , et al. 2015,  
Self-Supervised Online Learning of Basic Object 
Push Affordances[J]. International Journal of 
Advanced Robotic Systems, 12(24):45-53. 
[23] F  Zhang,  Li L , 2017, Research on Knowledge 
Push Method for Business Process Based on 
Multidimensional Hierarchical Context Model[J]. 
Jisuanji Fuzhu Sheji Yu Tuxingxue Xuebao/Journal 
of Computer-Aided Design and Computer 
Graphics29(4):751-758. 
[24] Hirchoua B ,  Ouhbi B ,  Frikh B . 2021.Deep 
Reinforcement Learning Based Trading Agents: Risk 
Curiosity Driven Learning for Financial Rules-Based 
Policy[J]. Expert Systems with Applications, , 
170(6):114-120. 
[25] Howard R ,  Zito C , 2021,. Learning Transferable 
Push Manipulation Skills in Novel Contexts[J]. 
Frontiers in Neurorobotics, 15(2):1-8. 
[26] Qian Y ,  Wang R ,  Wu J , et al, 2020, 
Reinforcement Learning Based Optimal Computing 
and Caching in Mobile Edge Network[J]. IEEE 
Journal on Selected Areas in Communications, 
4(99):1-10. 
[27] Zhu Z, Sun Y, 2023 Jul;27 Personalized information 
push system for education management based on big 
data mode and collaborative filtering algorithm. Soft 
Computing. (14):10057-67. 
[28]  Xu B. Artificial intelligence teaching system and 
data processing method based on big data. 
Complexity. 2021 May 5;2021:1-1.  
[29] Li Y, Meng S, Wang J. Research and application of 
personalized learning under the background of 
artificial intelligence. In2021 international conference 
on education, information management and service 
science (EIMSS) 2021 Jul 16 (pp. 54-57). IEEE. 
[30] Zhang Q. Construction of personalized learning 
platform based on collaborative filtering algorithm. 
Wireless Communications and Mobile Computing. 
2022 Mar 18;2022. 
[31] Zhao J, Wang Z, Jianjun Z. Construction on 
precise-personalized-learning evaluation system based 
on cipp evaluation model and integrated FCE-AHP 
method. Journal of Intelligent & Fuzzy Systems. 
2023(Preprint):1-3.  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

124   Informatica 48 (2024) 111-124                                                                        D. Wu