https://doi.org/10.31449/inf.v47i10.5174 Informatica 47 (2023) 79–92 79 
Construction of a Model for College Students' Innovation and 
Entrepreneurship Quality Based on Artificial Intelligence 
Technology 
Ren Zhiyi
*
, Zhao Nan, Shi Zhiyan
 
School of Foreign Languages, North China Institute of Aerospace Engineering, Langfang, 065099, China  
E-mail: renzhiyi@nciae.edu.cn, Zhiyihuo98@gmail.com, Sh.Gu01614314@gmail.com 
*Corresponding Author 
Keywords: artificial intelligence technology, big data collection, association rule features, data fusion, clustering 
algorithm, markov chain  
Received: September 10, 2023 
The quality of invention and entrepreneurship of university students is an important factor in building an 
innovative society in China. In order to improve the quality of innovation and entrepreneurship of college 
students, a model of college students' innovation and entrepreneurship quality based on artificial 
intelligence technology is constructed. The historical information data and real-time data streams of 
university students' invention and entrepreneurship quality are collected from data sources such as 
management departments and service and collaborative organizations. Data processing methods pre-
process the collected data and then fuse it with the collected data by association rule feature extraction 
methods. After the fusion, the components of the entrepreneurial quality model of college students are 
determined, 25 quality indicators are selected from five dimensions of innovation consciousness, 
entrepreneurship, emotional management, professional knowledge and practical ability to build a model 
of college students' innovation and entrepreneurship quality, and Markov chain and fuzzy algorithm are 
used to assess college students' invention and entrepreneurship quality. The results show that the model 
has an accuracy rate of more than 94% for feature extraction of association rules and a good clustering 
effect of indicators, which is able to enhance the invention and entrepreneurship quality of university 
students greatly and has a wide promotion value. 
Povzetek: Razvit je model umetne inteligence za izboljšanje kakovosti inovativnosti in podjetništva 
študentov, kar povečuje njihovo ustvarjalnost. 
 
1 Introduction 
Aiming to comprehend the rapid progress of the social 
economy, higher requirements are offered for the 
comprehensive quality and ability level of contemporary 
college students [1]. Innovation and entrepreneurship, as 
a new way of college students' employment [2], reflect 
the comprehensive competitiveness of a city and region 
[3], especially the most dynamic invention and 
entrepreneurship of university students is a new engine of 
urban economic development, and carrying out 
innovation and entrepreneurship has become a new way 
to solve the difficulty of college students' employment. 
Innovation and entrepreneurship drive the country's 
economic progression, bring about national growth, and 
have become a key element for countries to win the 
competition [4]. Firstly, innovation and entrepreneurship 
quality training can bring many inventive and 
entrepreneurial skills that will promote the upgrading of 
the national invention. Secondly, the innovative ideas and 
behaviors of talents with dual innovation quality will 
increase. They may generate more new solutions and 
project ideas, etc., to bring innovation and vitality to 
enterprises [5]. Then, cultivating college students with 
innovative and entrepreneurial qualities can promote 
universities to enhance their own vitality, increase  
 
research achievements, and achieve sustainable 
development. Finally, college students with a higher 
quality of innovation and entrepreneurship will develop 
themselves better [6]. Currently, as the economy is in 
continuous transformation, the economic characteristics 
of the nature of innovation and entrepreneurship are 
gradually becoming clear. At the same time, the lack of 
inventive and entrepreneurial skills in society and the 
imbalance between stock and demand make the 
importance of innovation and entrepreneurship 
increasingly obvious [ 7 ] . The quantity of talent cannot 
be used as a principle for judging the progress of society 
but rather focuses on the quality of such talent. University 
students are the core and basis of future invention and 
entrepreneurship, and not all of them have innovation and 
entrepreneurial potential, so the quality characteristics of 
innovation and entrepreneurial talents should be clarified, 
while the premise of identification and cultivation is to 
establish a quality model of college students' innovation 
and entrepreneurial talents. 
In the study of the problem of constructing the 
invention and entrepreneurship quality model of university 
students, used the iceberg model as the basis to 
systematically cultivate the knowledge structure of 
university students, improve their invention and 
entrepreneurship skills, help them create their core values 

80 Informatica 47 (2023) 79–92 R. Zhiyi et al. 
of innovation and entrepreneurship, form correct self-
cognition, shape their innovation and entrepreneurship 
qualities and stimulate their internal and external 
motivation [8 ] . The selected quality indicators are too 
idealized and detached from reality. Authors used multiple 
regression analysis ways to analyze the invention and 
entrepreneurship quality data based on the definition of 
innovation and entrepreneurship quality and constructed 
the regression model of innovation and entrepreneurship 
quality improvement, and found that the elements like 
invention and entrepreneurship teaching, cooperation and 
innovation between teachers and students, incentive 
system guarantee, and the creation of dual-innovation 
atmosphere have an important role in the improvement 
of innovation and entrepreneurship quality [ 9 ] . In the 
process of constructing this model, the information 
sources of college students' innovation and 
entrepreneurship quality lack timeliness and 
comprehensiveness, which causes the model created to be 
constantly updated and perfected and wasted resources. 
During the procedure of creating the invention and 
entrepreneurship quality model, Duan creatively divided 
the innovation and entrepreneurship process into seven 
stages and built a new innovation and entrepreneurship 
quality model. Then a system of innovation and 
entrepreneurship quality evaluation indexes containing 7 
secondary and 21 tertiary indicators was constructed, 
which can objectively and systematically reveal the 
growth logic of innovation [10]. The evaluation method 
used in constructing this model lacks relevance and 
personalization, which leads to the lack of fairness and 
predictability of the final evaluation results of college 
students' innovation and entrepreneurship quality. Ref 
[1 1 ] in their study of university students' invention and 
entrepreneurship quality, according to the competency 
model, build a quality model for entrepreneurship and 
entrepreneurship of university students. The entrepreneurial 
quality indicators selected in this model are one-sided, 
which affects the cultivation of innovation and 
entrepreneurship quality of college students. In view of 
the above problems, the model of college students' 
innovation and entrepreneurship quality based on 
artificial intelligence technology is constructed. It is 
hoped that the research results can be used to develop 
university students' invention and entrepreneurship 
quality cultivation system in a targeted way, which can 
provide a reference basis for major universities to 
formulate the corresponding curriculum system, 
examination objectives and employment guidance. 
2 Materials and methods 
2.1 Data collection and pre-processing of 
college students' invention and 
entrepreneurship 
The big data related to college students' invention and 
entrepreneurship quality are collected. The data 
processing methods in AI technology are used to 
complete the pre-processing operations of data denoising, 
integration, cleaning, normalization, transformation, 
discrimination, discretization, extraction, etc. The data 
sources of the great info on university students' 
invention and entrepreneurship quality are shown in 
Figure 1. The collected big data of college students' 
invention and entrepreneurship quality are fused and 
processed [12], and on this basis, correlation analysis, 
data mining and modelling analysis are carried out on the 
data; services are provided for the construction of college 
students' innovation and entrepreneurship quality model. 
 
Figure 1: Structure chart of big data collection sources of university students' invention and entrepreneurship quality 

Construction of a Model for College Students' Innovation and… Informatica 47 (2023) 79–92 81 
Aiming to ensure the completeness and diversity of 
the great info data on the innovation and 
entrepreneurship quality of university students, we 
should do everything possible to collect information 
about the invention and entrepreneurship quality of 
university students and call this information university 
students' invention and entrepreneurship quality great 
info. The data sources are very wide, and the 
completeness, comprehensiveness and diversity of data 
are very unusual and important issues in all big data 
problems [13], directly affecting the effect of subsequent 
data mining and data analysis. 
The sources of innovation and entrepreneurship 
quality data of college students include four aspects: 
innovation and entrepreneurship management department, 
innovation and entrepreneurship service and collaborative 
organizations, consumers of products and college 
students themselves. Among them, the information of the 
innovation and entrepreneurship management department 
includes various policy documents of the government 
departments, relevant management regulations issued by 
colleges and universities, investment in the relevant 
courses offered by colleges and universities, data 
information of the industry and commerce department, tax 
information of the tax department, innovation 
achievements information of the patent department, 
lawsuit information of the legal department, publicity and 
punishment information of the environmental protection 
department, etc.; Innovation and entrepreneurship services 
and cooperation organizations include various innovation 
and entrepreneurship competition documents, 
management and statistical data of incubation parks, 
reports of achievement transformation platforms, 
information on invention and entrepreneurship training for 
university students, investment of angel investors and 
partner investors, bank loan information, etc.; The 
consumer information of the corresponding product 
includes the consumer's evaluation and feedback 
information of the product, the repurchase rate of the 
product, consumer preferences, complaints and reporting 
information, etc.; The information from the innovation 
and entrepreneurship college students themselves is more 
diverse, including the awards of the invention and 
entrepreneurship contest, personal credit, income 
information, course learning, consumption and shopping, 
loans and financing, etc. 
After collecting the university students' invention and 
entrepreneurship quality big data in each data source and 
completing the pre-processing, the association rules 
within the university students' invention and 
entrepreneurship quality great info are extracted using the 
association rule factor derivation way in artificial 
intelligence technology [14]. According to this, the great 
info phase space dispensation 𝛷 of college students' 
innovation and entrepreneurship quality evaluation is 
made, which can be perceived as a 𝑛 ×𝑚 data control 
matrix of university students' invention and 
entrepreneurship quality. With 𝜕 𝑞 and 𝑃 (𝑛 𝑖 )=
{𝜕 𝑘 |𝜕 𝑟 𝑘𝑗
=1,𝑘 =1,2,⋯,𝑚 } denoting the detail 
distribution vector and possibility dispensation function, 
respectively, the fusion assessment of university students' 
invention and entrepreneurship quality data is realized. 
Analyze association rule feature data [ 1 5 ] , and 
the regression analysis model of the big data of 
college students' innovation and entrepreneurship 
quality is obtained. The fuzzy affiliation function for 
identifying college students' innovation and 
entrepreneurship quality big data is constructed, and 
the formula is described as follows. 
𝐻 𝑠 (0)
=∑⟨𝐻 𝑠 (𝑛 )
,𝑦 𝛾𝑛
⟩
𝑘 𝑛 =0
𝑦 𝛾𝑛
+𝐻 𝑠 (𝑘 +1)
 (1) 
In equation (1), 𝐻 𝑠 (𝑛 )
 represents the characteristic 
quantity of big data distribution of college students' 
innovation and entrepreneurship quality; 𝑦 𝛾𝑛
 represents 
the number of data reconstruction bits; 𝐻 𝑠 (𝑘 +1)
 represents 
the iterative coefficient of the evaluation of university 
students' invention and entrepreneurship quality. 
According to the above resolution, the convertible  
decomposition formula of the great info of  
university students' invention and  
 
( ) ( ) ( )  
( ) ( ) ( )

( )

12
min
min
m
K
R t R t R t
F t sign k t w T t



=+


= −  + 
 


 
 
entrepreneurship quality is determined as follows.  
In equation (2), 𝑘 𝜇 (𝑡 ) and 𝑤 denote the sampling 
scale and relative weight of college students' innovation 
and entrepreneurship quality data at 𝑡 , 𝛥 𝑇 𝑚 (𝑡 ) and 𝛩 𝛩 
show the quantitative detail group of college students' 
innovation and entrepreneurship quality data at 𝑡 and the 
adaptable possibility term of 𝑘 𝜇 (𝑡 ). 
2.2 Clustering of quality indicators based 
on artificial intelligence technology 
After completing the big data fusion of college students' 
entrepreneurial quality, the clustering algorithm in 
artificial intelligence technology is used to determine the 
constituents of college students' entrepreneurial quality 
model for the fused data [16]. 
Among various clustering algorithms, the K-means 
algorithm has become the most widely used clustering 
algorithm [17]. The algorithm has a certain processing 
capacity for large-scale data sets, consisting of two stages: 
initialization and iteration. The initialization phase 
randomly assigns the initial class cluster centres. The 
iterative phase has two steps; the first step (also called the 
assignment step) places all data objects into the class 
clusters closest to each centroid if the initial class cluster 
centres are available; the second step (also called the 
update step) updates the centroids of all class clusters. In 
the classical K-means algorithm, its initial class cluster 
centres are generated using a random method [18]. An 
unreasonable initial class cluster centre will largely reduce 
(2) 

82 Informatica 47 (2023) 79–92 R. Zhiyi et al. 
the clustering quality of the K-means algorithm. At the 
same time, improper initialization will also increase the 
running time of the K-means algorithm. 
On the other hand, the update step of the K-means 
algorithm is also important for the performance 
improvement of the K-means algorithm. In general, the 
algorithm requires several adjustments of the class cluster 
centres before reaching acceptable clustering results. This 
step updates the corresponding centroids by calculating 
the average distances of all data objects in each class 
cluster. However, the process of forming new class cluster 
centres is vulnerable to the interference of outliers. 
Improper outlier handling can lead to large deviations 
between the new and actual centres; thus, the 
corresponding class clusters are "distorted" by the outliers. 
Based on this, an improved K-means algorithm based on 
density parameter and centre replacement (DC-Kmeans 
algorithm for short) is proposed. 
The DC-Kmeans algorithm first determines the 
initial class cluster centres (main dimensional quality 
indicators) by estimating the density aspect of every info 
item in the fused college student innovation and 
entrepreneurship quality dataset [19] so as to avoid the 
problem of unsteady grouping results made by random 
selection of the primary class group centres. Furthermore, 
the class group centres are replaced by the bias created by 
the traditional K-means method to prevent the impact of 
outliers on the grouping outcomes. 
2.2.1 Selection of initial class cluster centres based 
on density parameters 
The DC-Kmeans algorithm pursues the method of gradual 
picking of class cluster centres on the basis of the density 
parameter. The following assumptions are set: in the 
Euclidean space 𝑅 𝑚 , the fused college students' 
innovation and entrepreneurship quality dataset 𝐵 =
{𝑥 1
,𝑥 2
,⋯,𝑥 𝑛 } contains 𝑛 data objects, and each object 
𝑥 𝑖 ={𝑥 𝑖 1
,𝑥 𝑖 2
,⋯,𝑥 𝑖𝑚
} has 𝑚 attributes. At the same time, 
the fused student innovation and entrepreneurship quality 
dataset 𝐵 is divided into 𝐾 clusters 𝐿 ={𝐿 1
,𝐿 2
,⋯,𝐿 𝐾 } 
using a clustering method. Where |𝐿 𝑘 | is the amount of 
info items included in the class cluster 𝐿 𝑘 ; the centroid of 
every class cluster in the group of class clusters 𝐿 is 𝑉 =
{𝑣 1
,𝑣 2
,⋯,𝑣 𝐾 } . The Euclidean distance d between any 
two data objects 𝑥 𝑖 and 𝑥 𝑗 (𝑖 ,𝑗 =1,2,⋯,𝑛 ) in the data set 
𝐵 is defined as: 
𝑑 (𝑥 𝑖 ,𝑥 𝑗 )
=
√
(𝑥 𝑖 1
−𝑥 𝑗 1
)
2
+(𝑥 𝑖 2
−𝑥 𝑗 2
)
2
+⋯+(𝑥 𝑖𝑚
−𝑥 𝑗𝑚
)
2
 
(3) 
 
According to the Euclidean distance, the biggest 
distance (𝐿 𝑎 ) and the shortest distance (𝑆 𝑚 ) among every 
info item are indicated as shown below: 
{
 
 
 
 
𝐿 𝑎 =∑ 𝑚𝑎𝑥 1≤𝑖 <𝑗 ≤𝑛 𝑑 (𝑥 𝑖 ,𝑥 𝑗 )
2
𝑛 −1
𝑖 =1
𝑆 𝑚 =∑ 𝑚𝑖𝑛 1≤𝑖 <𝑗 ≤𝑛 𝑑 (𝑥 𝑖 ,𝑥 𝑗 )
2
𝑛 −1
𝑖 =1
 (4) 
Even though it is considered that the dataset 𝐵 is 
parted into K class clusters, the number of info items in 
every class cluster created may be distinct for different 
clustering methods. By changing the number of 
information items, the gap among every information item 
pair also changes [20]. For this purpose, the dynamic 
medium distance (𝐷 𝑦 ) on the basis of the most and least 
distances among every information item is defined as: 
𝐷 𝑦 =
𝐿 𝑎 +𝑆 𝑚 2𝐾 (5) 
Within Eq. (5), K is the number of class clusters into 
which the data set 𝐵 is divided. 
In the process of finding the initial class cluster 
centres, most density-based clustering methods mostly 
rely on exterior principles for the selection of the 
neighborhood radius [21], which means that the 
algorithm's performance can be significantly affected by 
improper parameter selection. To address this problem, a 
dynamic medium distance (𝐷 𝑦 ) on the basis of the most 
and least distances among every information item is first 
defined. The distance changes dynamically with every 
iteration process. The neighbourhood radius of distinct 
grouping steps can be obtained in time to identify more 
efficient and stable density principles and decrease the 
impact of the exterior principles on the clustering 
outcomes. 
The data object with the highest density parameter is 
the first initial class cluster centre and is removed from the 
original dataset 𝐵 . At the same time, the data objects in a 
circular area centred on the first primary class cluster 
centre with radius 𝐷 𝑦 are removed from 𝐵 . The second 
initial class cluster centre is the data object with the 
highest remaining density parameter in the data set 𝐵 . The 
second initial class cluster centre is the data object with 
the highest density parameter in the data set 𝐵 . This 
continues until the specified K initial class cluster centres 
are found. 
2.2.2 Replacement of class cluster centres 
Another drawback of the traditional K-means method is 
that it is very sensitive to outliers in the data set. Actually, 
some of the primary class centres created by the traditional 
K-means method are not the true class centres in the 
aiming data set (these factors are called "pseudo centres"). 
Additionally, due to the influence of outliers, the positions 
of the created class centres may digress from the real class 
centres [22]. This error will greatly decrease the precision 
of the traditional K-means method. 
The centres created by the K-medoids clustering 
method are the true information points of the aiming data 

Construction of a Model for College Students' Innovation and… Informatica 47 (2023) 79–92 83 
set all the time. Influenced by the K-medoids method, a 
centre replacement method is proposed to upgrade the 
pseudo centres created by the traditional K-means method. 
Specifically, by the time the K-means method generates a 
pseudo centre for a class cluster, it is exchanged by the 
closest neighbouring point in that class cluster. While that 
neighbouring point must be as far as possible from the 
outliers of that class cluster. During the clustering 
procedure, the pseudo-class cluster centers are upgraded 
sequentially until every true class cluster centre is 
determined. 
2.2.3 Flow of DC-Kmeans algorithm 
The procedure of clustering university students' invention 
and entrepreneurship quality factors based on the DC-
Kmeans method is represented in Figure 2. DC-Kmeans 
method is not only able to find the primary class cluster 
centres persistently but also has the ability to handle 
outliers. In the DC-Kmeans algorithm, step (1) calculates 
the dynamic average distance among all university 
students' invention and entrepreneurship quality data 
objects in data set 𝐵 . Step (2) calculates the density 
parameters of all university students' invention and 
entrepreneurship quality data objects. Step (3) finds the K 
initial class cluster centres of college students' innovation 
and entrepreneurship quality data set 𝐵 and puts them into 
set V based on the density principle. Steps (4) to (8) figure 
out the final partitioning of college students' innovation 
and entrepreneurship quality data set 𝐵 . Specifically, step 
(5) initializes each class cluster; step (6) puts the 
information objects into the respective class clusters; step 
(7) updates the class cluster centres using the centre 
replacement method. 
 
 
Figure 2: Clustering process of university students' invention and entrepreneurship quality features according to the 
DC-Kmeans algorithm 
In the clustering process based on the DC-Kmeans 
algorithm, the input is the data set of college students' 
innovation and entrepreneurship quality 𝐵 =
{𝑥 1
,𝑥 𝑖 ,⋯,𝑥 𝑛 } and the number of class clusters 𝐾 , and the 
input is the dataset 𝐿 ={𝐿 1
,𝐿 2
,⋯,𝐿 𝐾 }. 
(1) Estimate the dynamic medium distance 𝐷 𝑦 among 
any pair of information items (𝑥 𝑖 ,𝑥 𝑗 ) in the university 
students' invention and entrepreneurship quality dataset 𝐵 . 
(2) Estimate the density parameter 𝜌 (𝑥 𝑖 ,𝐷 𝑦 ) of the 
information item 𝑥 𝑖 . 
(3) Select the data object x with the highest density 
parameter from the student innovation and 
entrepreneurship quality dataset 𝐵 , and remove all data 
objects in the neighbourhood of x from the student 
innovation and entrepreneurship quality dataset 𝐵 ; set x as 
the k-th initial class cluster centre, as 𝑣 𝑘 ; put 𝑣 𝑘 into the 
set 𝑉 of primary class cluster centres. 
(4) Loop the process (1) to (3). 
(5) Initialize the different class clusters using 
equation (7). 

84 Informatica 47 (2023) 79–92 R. Zhiyi et al. 
𝐿𝑒𝑡 𝐿 𝑘 =𝜙 (1≤𝑘 ≤𝐾 ) (6) 
(6) Estimate the distance among the centres of every 
class cluster of each information object 𝑥 𝑖 and 𝑉 in the 
data set 𝐵 , and put 𝑥 𝑖 into the related class cluster based 
on the closest principle. 
(7) Estimate the distance between the centre 𝑣 𝑘 of the 
class cluster 𝐿 𝐾 and the rest of the information items in the 
class cluster, find the information items 𝑣 𝑘 ′
 that is closest 
to 𝑣 𝑘 , while 𝑣 𝑘 ′
 should be as far as possible from the class 
group point in 𝐿 𝐾 , and update the new centre of the class 
cluster 𝐿 𝐾 . 
 
 
 
(8) To be the standard function ∑ ∑ 𝑑 (𝑣 𝑖 ,𝑥 )
2
𝑋 ∈𝐶 𝑖 𝐾 𝑖 =1
 
converges to a constant, under which 𝑣 𝑘 is the new centre 
of the class cluster 𝐿 𝐾 . 
2.3 Construction of college students' 
innovation and entrepreneurship 
quality model 
After the above series of basic research, a five-
dimensional quality model of college students' innovation 
and entrepreneurship is established, as shown in Figure 3, 
which includes five dimensions and 25 sub-dimensions of 
innovation consciousness, entrepreneurial spirit, 
emotional management, professional knowledge and 
practical ability. 
 
Figure 3: Innovation and entrepreneurship quality model of college students 
The interpretation of each main dimension is as 
follows. 
(1) Innovation consciousness: it refers to some 
psychological characteristics and competence traits that 
must be possessed to engage in innovative activities [23].  
(2) Entrepreneurship: it refers to some ability traits 
and psychological characteristics that must be possessed 
to engage in entrepreneurial activities.  
(3) Emotion management: it refers to the individual's 
emotional resilience in the face of negative emotions such 

Construction of a Model for College Students' Innovation and… Informatica 47 (2023) 79–92 85 
as failure, anxiety, and stress, as well as the ability to cope 
[24]. 
(4) Professional knowledge: it refers to the 
professional knowledge necessary to engage in innovative 
and entrepreneurial activities.  
(5) Practical ability: it refers to some abilities and 
traits required to engage in innovative and entrepreneurial 
activities.  
After determining college students' innovation and 
entrepreneurship quality model, the Markov chain and 
fuzzy algorithm in artificial intelligence technology are 
used to construct the evaluation model of college students' 
innovation and entrepreneurship quality [25]. The specific 
process is as follows. 
(1) Invite six experts in the fields related to 
innovation and entrepreneurship of college students to 
form a judging group. 
(2) Construct the evaluation scale set. 
 
𝐶 =[
Excellent 、Good 、 
 Moderate 、commonly 、 Poor
] (7) 
 
The scores corresponding to different evaluation 
scales 𝐶 𝑖 in equation (8) are 0.9, 0.7, 0.5, 0.3 and 0.1, 
respectively. 
Let the grade of a certain quality of college students' 
innovation and entrepreneurship be evaluated 
corresponding to the set of evaluation scales, which is 
expressed as 𝑖 =1,2,3,4,5, and the probability of 
conversion of this index from grade 𝑖 to 𝑗 can be expressed 
by the transfer probability 𝑍 𝑖𝑗
. 
The state vector is used to describe the proportion of 
a quality indicator of university students' invention and 
entrepreneurship quality for evaluation level 𝑖 , which is 
expressed as 𝑆 (0)
=[𝑠 1
,𝑠 2
,⋯,𝑠 5
], where 𝑠 𝑖 is the 
proportion of a quality indicator of college students' 
innovation and entrepreneurship quality in evaluation 
level 𝑖 . Suppose the evaluation of college students' 
innovation and entrepreneurship quality is implemented 
once every three months. After three months, the change 
in attitude can be defined as a one-step transfer. In that 
case, six experts are identified to score different quality 
indicators of university students' invention and 
entrepreneurship quality for six months so that the change 
in evaluation levels of different quality indicators can be 
obtained. The number of transfers of any quality indicator 
from level 𝑖 to level 𝑗 can be described by 𝑛 𝑖𝑗
. According 
to Markov chain theory, it can use a one-step transfer 
probability matrix 𝑍 to describe this transfer phenomenon. 
This probability is the one-step transfer probability of a 
Markov chain, which can be used to reach a steady state 
after several transfers, i.e., the result is fixed during the 
subsequent transformation. It can be described by the 
steady state vector 𝑈 =[𝑢 1
,𝑢 2
,⋯,𝑢 5
], which can be 
solved by equation (8) to obtain. 
{
𝑈 =𝑈𝑍
∑𝑢 𝑖 5
𝑖 =1
=1
 (8) 
Based on the above process, it is possible to 
determine the stable vector U
k
 of the states of different 
college students' innovation and entrepreneurship quality 
indicators. 
If the fuzzy correlation from the set of quality 
indicators to the set of evaluation scales can be described 
by the affiliation matrix β [26], then β
ki
 can be used to 
describe the probability level of getting the i th evaluation 
of the innovation and entrepreneurship quality of college 
students by k quality indicators, and β is described as 
follows. 
𝛽 =[𝑈 1
,𝑈 2
,⋯,𝑈 𝑛 ]
𝑇 
(9) 
(3) The set of weights is denoted by 𝑊 =
[𝑤 1
,𝑤 2
,⋯,𝑤 𝑛 ], where 𝑤 has been determined from the 
hierarchical analysis, from which a comprehensive rating 
vector E is obtained, described by the following 
equation. 
𝐸 =𝑊 ∗𝛽 =[𝑒 1
,𝑒 2
,⋯,𝑒 5
] (10) 
In equation (10), ∗ and 𝑒 𝑖 denote the matrix 
multiplication operation and the score probability of 
university students' invention and entrepreneurship 
quality index corresponding to 𝐶 𝑖 , respectively. 
Equation (11) is used to calculate the invention and 
entrepreneurship quality evaluation value of university 
students. 
𝑁 =𝐸 [0.9 0.7 0.5 0.3 0.1]
𝑇 (11) 
In equation (11), 𝑁 indicates the evaluation result of 
college students' innovation and entrepreneurship quality. 
3 Experimental results 
To verify the application effect of the model built in this 
paper in the actual application process, some students 
from six colleges and universities in a city are selected as 
the research subjects. The model of this paper is used to 
cultivate the innovation and entrepreneurship quality of 
the research subjects. The total number of research 
subjects is 1500, mainly students majoring in finance, 
credit management, investment and trade economics in the 
selected universities, among which the number of male 
and female students are 793 and 707, respectively. 
3.1 Feature extraction test of association 
rule  
The accuracy of feature extraction of association rules in 
the innovation and entrepreneurship quality model of 
college students is tested. The test results of association 
rule feature extraction precision of the model in this article 
under different data volume conditions are shown in 
Figure 4. 

86 Informatica 47 (2023) 79–92 R. Zhiyi et al. 
 
Figure 4: Test outcomes of the accuracy of feature extraction of association rules in this model 
Analyzing Figure 4, we can see that the decision rate 
of association rule features extracted by the model of this 
paper in the process of great information fusion of college 
students' innovation and entrepreneurship quality 
indicates a downward trend under the term that the 
quantity of university students' invention and 
entrepreneurship quality big data gradually increases. 
Under the condition that the quantity of big data of 
university students' invention and entrepreneurship 
quality is small, the accuracy rate of association rule 
features extracted by the model in this paper is high; with 
the gradual increase of data quantity, the accuracy rate of 
association rule features decreases gradually. When the 
data volume reaches 10GB, the precision rate of the 
association rule features extracted by this model is above 
94%. The above results fully demonstrate that the model 
of this paper can accurately extract the association rules in 
the process of fusion of university students' invention and 
entrepreneurship quality data, which is beneficial to the 
determination of college students' invention and 
entrepreneurship quality index. 
3.2 Homogeneity and completeness test of 
quality indexes 
In order to verify the clustering performance of the model 
in this paper, homogeneity and completeness are used as 
the test indexes. These two indicators are usually used 
together, and the values of both test indicators range from 
0 to 1. The larger the value is, the better the clustering 
effect is. Figure 5 shows the homogeneity and 
completeness test results of the clustering results of 
college students' innovation and entrepreneurship quality 
indicators in this paper. 
 
Figure 5: Homogeneity and integrity test results of clustering results of quality indicators in the model in this paper 
  

Construction of a Model for College Students' Innovation and… Informatica 47 (2023) 79–92 87 
Analyzing Figure 5, we can get that the homogeneity 
value of each main dimension of quality indicators reaches 
above 0.75 in the process of clustering the innovation and 
entrepreneurship quality indicators of college students 
using the model in this paper, which indicates that the 
clustering results of indicators obtained from this model 
have good homogeneity. At the same time, the 
completeness value of each main dimension quality index 
reaches 0.7 or more, which indicates that the clustering 
results of the indexes obtained from this model have good 
homogeneity and completeness. 
To further verify the indicator clustering effect of the 
model in this paper, the model of reference [8], the model 
of reference [9], the model of reference [10] and the model 
of reference [11] are used as comparison models, and the 
adjusted Rand coefficient, mutual information and 
Fowlkes-Mallow’s index are used as comparison 
indicators to analyze the indicator clustering performance 
of the model in this paper, and the comparison models and 
the obtained outcomes are presented in Figure 6. 
 
(a) Adjust the Rand coefficient 
 
(b) Mutual information 

88 Informatica 47 (2023) 79–92 R. Zhiyi et al. 
 
(c)Fowlkes-Mallow’s index 
Figure 6: Comparison results of clustering performance between the model in this paper and the comparison model 
The adjusted Rand coefficient describes the 
similarity between the distributions of two data (quality 
indicators), and its value ranges from -1 to +1. The larger 
the value is, the more consistent the clustering results are 
with the actual situation. If the value is negative, it means 
that the two data distributions are independent of each 
other, and the degree of matching is low. The mutual 
information and the Fowlkes-Mallows index can describe 
whether there is a correlation between two variables 
(quality indicators) and how close the relationship is. The 
values of mutual information and the Fowlkes-Mallows 
index are from 0 to 1, and the larger the value is, the closer 
the relationship between the two variables is. The analysis 
of Figure 6 shows that in the clustering process of this 
model, the adjusted Rand coefficients are above 0.75, and 
the mutual information and Fowlkes-Mallow’s indices are 
above 0.7 and 0.75, respectively, which show significant 
advantages compared with the four comparison models. 
Based on the above data, combined with the results in 
Figure 5, it can fully illustrate that this paper's model has 
a better clustering effect in clustering college students' 
innovation and entrepreneurship quality indicators. 
3.3 Evaluation results of 
entrepreneurship and entrepreneurial 
quality 
The model of this paper is used to evaluate the innovation 
and entrepreneurship quality of the four majors, and the 
obtained results are shown in Figure 7. 
 
Figure 7: Evaluation results 
 
 
 
 
 
 

Construction of a Model for College Students' Innovation and… Informatica 47 (2023) 79–92 89 
 
Analyzing Figure 7, we get that among the four 
majors, the innovation and entrepreneurship quality of 
students in finance, credit management and trade 
economics are in the better grade, and the innovation and 
entrepreneurship quality of students in investment science 
is in the average grade. Based on the model of this paper, 
the innovative and entrepreneurial quality of the students 
of the four majors is specially cultivated, and the 
innovative and entrepreneurial quality of the students of 
the four majors is evaluated again after one year and 
compared with the first evaluation results, and the 
obtained outcomes are presented in Figure 8. 
 
Figure 8: Comparison results of innovation and entrepreneurship quality evaluation 
 
Analysis of Figure 8 shows that the invention and 
entrepreneurship quality evaluation value of students in all 
four majors has increased to different degrees, among 
which the innovation and entrepreneurship quality 
evaluation value of students in finance and investment 
majors shows a certain upward trend, while the innovation 
and entrepreneurship quality evaluation grade of students 
in credit management and trade economics has increased 
by one level. This can show that the model based on this 
paper can significantly improve the quality of invention 
and entrepreneurship of university students and has 
extensive promotion value. 
4 Discussion 
The cultivation of college students with innovative and 
entrepreneurial qualities cannot be shaped by slogans or 
speeches alone. To cultivate university students with 
invention qualities, we need to rely on professional 
cultivation programs, make innovative and 
entrepreneurial literacy an essential educational goal, 
develop relevant resources and curriculum arrangements 
for innovation and entrepreneurship, and continuously 
practice and transform them. 
4.1 Cultivation of explicit qualities - 
knowledge and ability 
For the explicit qualities of innovation and 
entrepreneurship quality of college students--knowledge 
and ability, colleges and universities can cultivate them 
through the following ways.  
4.1.1 Systematic cultivation of knowledge 
structure of college students  
(1) Rational construction of curriculum system. 
Integrating the concept and requirements of innovation 
and entrepreneurship into the professional curriculum and 
teaching is important. It leads students to explore and 
understand the industry and market prospects and trends, 
stimulating students' interest in invention and 
entrepreneurship and cultivating their invention and 
entrepreneurship skills in the actual teaching. 
(2) Scientific integration of faculty. The construction 
of faculty is very important for the cultivation of 
knowledge. Colleges and universities should enrich the 
faculty of innovation and entrepreneurship quality training 
as soon as possible and increase investment, not only to 
strengthen the training of existing lecturers but also to hire 
entrepreneurs, business executives, and heads of relevant 
government departments as special teachers to teach 
students. General education teachers, professional course 
teachers and innovation and entrepreneurship instructors 
should work together to break through knowledge barriers 
in the three areas and help students establish a systematic 
knowledge framework. At the same time, innovative 
teaching methods should be used to reject rigidity and 
stiffness and promote college students' knowledge 
comprehension.  
4.1.2 Improvement of innovation and 
entrepreneurial skills  
(1) Reasonable arrangement of practical activities. In 
order to create an atmosphere of innovation and 
entrepreneurship for students, colleges and universities 

90 Informatica 47 (2023) 79–92 R. Zhiyi et al. 
can hold innovation and entrepreneurship competitions 
regularly, allowing students to present their project 
products through group or individual participation, create 
product plans and present them in the form of PPT, and 
invite professional teachers and relevant persons in charge 
of cooperative enterprises as judges to guide students' 
entrepreneurial ideas. 
(2) Establish a sound innovation and 
entrepreneurship platform. Teachers lead students to visit 
creative workshops of off-campus cooperative enterprises, 
connect entrepreneurs with targeted off-campus practice 
platforms, truly experience the cultural atmosphere of 
entrepreneurial enterprises, and participate in various 
innovation and entrepreneurship activities organized by 
off-campus practice platforms, such as project roadshow 
activities, investment and financing matching activities, 
entrepreneurship salons, etc., so as to open up the horizons 
of college students and help those with entrepreneurial 
ideas to move from entrepreneurial training to 
entrepreneurial practice. 
(3) Focus on learning ability. Colleges and 
universities should create innovations in teaching mode, 
focus on cultivating college students' learning abilities and 
strengthen their own learning motivation. 
(4) Pay attention to cultivating resistance to stress 
and the ability to withstand setbacks. It should further 
stimulate students' desire for knowledge and innovative 
thinking, strengthen the cultivation of college students' 
willpower, correctly guide their emotional adjustment 
ability, and appropriately improve their anti-stress ability. 
4.2 Cultivation of implicit qualities--
innovation consciousness, 
entrepreneurship, emotional 
management 
Because they cannot be directly measured and quickly 
cultivated, the implicit qualities below the iceberg often 
do not receive as much attention as the explicit qualities. 
Still, these implicit qualities are the key factors affecting 
college students' innovation and entrepreneurship. 
Colleges and universities can cultivate these hidden 
qualities in subtle ways. 
(1) Establishing core values of innovation and 
entrepreneurship 
The establishment of values does not happen 
overnight but is a process of gradual accumulation. 
Colleges and universities can strengthen innovation and 
entrepreneurship role model education. By 
communicating with role models, university students can 
deepen their understanding of invention and 
entrepreneurship and take the invention to learn the 
shining points of role models. Colleges and universities 
should create a positive atmosphere of innovation and 
entrepreneurship. 
(2) Guiding correct self-perception 
Research shows that good self-awareness is the 
primary factor in determining leadership influence, so 
colleges and universities should do the following. 
1) Carry out self-cognition courses to make college 
students look directly inside to form self-knowledge. 
2) Provide scientific psychological guidance so 
college students' self-understanding will not deviate from 
the track. On the basis of correct self-understanding, they 
should become confident and optimistic people. 
(3) Shaping innovative and entrepreneurial qualities 
Innovative and entrepreneurial quality is one of the 
key factors to determine the success of college students' 
innovative and entrepreneurial activities, which is difficult 
to change in the short term and can only be cultivated 
carefully in the long term: holding regular learning 
activities with different themes, carrying out practical 
expansion activities, focusing on cultivating invention and 
innovation, and focusing on the cultivation of 
responsibility. 
(4) Stimulating the internal and external motivation 
of university students 
1) Colleges and universities can guide college 
students to set innovation and entrepreneurship goals 
through teaching courses, daily activities and teachers' 
guidance, and encourage them to take innovation and 
entrepreneurship as one of the main ways to realize their 
self-worth. 
2) Colleges and universities can consider students in 
various aspects according to their plans and operability 
and implement incentive mechanisms. Schools can build 
entrepreneurial streets, combine the plans submitted by 
students to lease their stores and give them certain 
incentives according to the turnover of the stores for 
college students to start their own businesses so that 
students can generate more enthusiasm in the innovation 
process and entrepreneurship. 
5 Conclusion 
The cultivation of innovation and entrepreneurship quality 
is the key to deepening the reform of university education. 
It is a great event for individuals, society and even the 
country for college students to engage in innovation and 
entrepreneurship. The employment pressure of college 
students can be solved by cultivating the innovation and 
entrepreneurship quality of college students. Therefore, 
establishing student innovation and entrepreneurship 
quality `models has important theoretical and practical 
significance. This article constructs a model of students' 
invention and entrepreneurship quality on the basis of 
artificial intelligence technology, hoping to deepen the 
understanding of students' innovation and 
entrepreneurship talent quality through the model, provide 
correct cognition and methods for the cultivation of 
innovation and entrepreneurship talents, and thus set up 
targeted cultivation programs for universities and provide 
guidance for enterprises to recruit relevant talents. 
Data Availability 
The raw data supporting the conclusions of this article wi
ll be made available by the authors, without undue reserv
ation." 
 
 

Construction of a Model for College Students' Innovation and… Informatica 47 (2023) 79–92 91 
Conflicts of Interest 
The authors declared that they have no conflicts of intere
st regarding this work." 
Authorship Contribution Statement 
Ren Zhiyi: Writing-Original draft preparation, 
Conceptualization, Supervision, Project administration. 
Zhao Nan: Methodology, Software, Validation 
Shi Zhiyan: Writing-Original draft preparation 
Refrences 
[1] Su X, Liu S, Zhang S, Liu L (2020) To be happy: 
a case study of entrepreneurial motivation and 
entrepreneurial process from the perspective of 
positive psychology. Sustainability 12:584. 
https://doi.org/10.3390/su12020584 
[2] Yuan C-H, Wang D, Su H-H (2023) Red Boat 
Spirit and Social Entrepreneurship Education in 
China. Sustainability 15:5030. 
https://doi.org/10.3390/su15065030 
[3] Srivastava S, Panda S, Williams Jr WA (2022) 
User-entrepreneurship and innovation: does 
customer involvement and learning orientation 
matter? Journal of Business & Industrial 
Marketing 37:1182–1196. 
https://doi.org/10.1108/JBIM-05-2021-0255 
[4] Schaeffer PR, Guerrero M, Fischer BB (2021) 
Mutualism in ecosystems of innovation and 
entrepreneurship: A bidirectional perspective on 
universities’ linkages. J Bus Res 134:184–197. 
https://doi.org/10.1016/j.jbusres.2021.05.039 
[5] LIU Y, HAO X, LI X (2020) Narcissistic 
personality and change in crucial entrepreneurial 
behavior in the context of serial entrepreneurship. 
Advances in Psychological Science 28:1083. doi: 
10.3724/SP.J.1042.2020.01083 
[6] Ionescu GH, Firoiu D, Pîrvu R, et al (2020) The 
potential for innovation and entrepreneurship in 
EU countries in the context of sustainable 
development. Sustainability 12:7250. 
https://doi.org/10.3390/su12187250 
[7] Slišāne A, Lāma G, Rubene Z (2021) Self-
assessment of the entrepreneurial competence of 
teacher education students in the remote study 
process. Sustainability 13:6424. 
https://doi.org/10.3390/su13116424 
[8] Wang Y (2022) Optimizing the cultivation path of 
college students’ innovation and entrepreneurship 
ability from the perspective of the internet. Wirel 
Commun Mob Comput 2022: 
https://doi.org/10.1155/2022/7973504 
[9] Yao D, He J, Yang W, Zhang M (2022) A Study 
on the Evaluation Index System of Innovation and 
Entrepreneurship Education for Undergraduate 
Students Majoring in Interdisciplinary Arts and 
Sciences. Higher Education Studies 12:135–145 
[10] Zhang F, Xi L (2021) An evaluation model for the 
innovation and entrepreneurship thinking ability 
of college students based on neural network. 
International Journal of Emerging Technologies 
in Learning (iJET) 16:188–204. 
https://doi.org/10.3991/ijet.v16i02.19729.  
[11] Niu B, Liu Q, Chen Y (2019) Research on the 
university innovation and entrepreneurship 
education comprehensive evaluation based on 
AHP method. International Journal of 
Information and Education Technology 9:623–
628. Doi: 10.18178/ijiet.2019.9.9.1278 
[12] Cao X, Guo S, Lin J, et al (2020) Online tracking 
of ants based on deep association metrics: method, 
dataset and evaluation. Pattern Recognit 
103:107233. 
https://doi.org/10.1016/j.patcog.2020.107233 
[13] Su Y, Zhu Z, Chen J, et al (2021) Factors 
influencing entrepreneurial intention of university 
students in China: integrating the perceived 
university support and theory of planned behavior. 
Sustainability13:4519. 
https://doi.org/10.3390/su13084519 
[14] Haller T, Völker C, Hartmann T (2020) Machine 
learning based multi-sensor fusion for the 
nondestructive testing of corrosion in concrete. 
The e-journal of nondestructive testing & 
ultrasonics 25:24890–24891 
[15] Chao Y, Xinian L, Xiaobin L (2022) A Dynamic 
Cloud Bayes Network-Based Cleaning Method of 
Multi-Source Unstructured Data. In: International 
Manufacturing Science and Engineering 
Conference. American Society of Mechanical 
Engineers, p V002T06A026. 
https://doi.org/10.1115/MSEC2022-85769 
[16] Ahmad S (2020) Adaptation of artificial 
intelligence in big data mining and its impact: a 
study. Solid State Technol 63:2322 
[17] Zhao K, Dai Y, Jia Z, Ji Y (2021) General fuzzy 
C-means clustering algorithm using Minkowski 
metric. Signal Processing 188:108161. 
https://doi.org/10.1016/j.sigpro.2021.108161 
[18] Ahmad A, Khan SS (2021) initKmix-A novel 
initial partition generation algorithm for clustering 
mixed data using k-means-based clustering. 
Expert Syst Appl 167:114149. 
https://doi.org/10.1016/j.eswa.2020.114149 
[19] Qin X, Li J, Hu W, Yang J (2020) Machine 
learning K-means clustering algorithm for 
interpolative separable density fitting to accelerate 
hybrid functional calculations with numerical 
atomic orbitals. J Phys Chem A 124:10066–
10074. https://doi.org/10.1021/acs.jpca.0c06019 
[20] Cai J, Liao D, Chen J, et al (2020) Research on 
data security protection method based on 
improved K-means clustering algorithm. In: 
Proceedings of the 4th International Conference 
on Big Data Research. pp 7–11. 
https://doi.org/10.1145/3445945.3445947 
[21] Khang TD, Vuong ND, Tran M-K, Fowler M 
(2020) Fuzzy C-means clustering algorithm with 
multiple fuzzification coefficients. Algorithms 
13:158. https://doi.org/10.3390/a13070158 

92 Informatica 47 (2023) 79–92 R. Zhiyi et al. 
[22] Liu Y, Li B (2020) Bayesian hierarchical K-means 
clustering. Intelligent Data Analysis 24:977–992. 
Doi: 10.3233/IDA-194807 
[23] Akhmetshin EM, Mueller JE, Yumashev A V, et 
al (2019) Acquisition of entrepreneurial skills and 
competences: Curriculum development and 
evaluation for higher education. J Entrep Educ 
22:1–12 
[24] Aleksin G, Kalbakk-Bøhler SC (2022) 
Environmental entrepreneurship and inclusive 
growth: A three-fold approach to analysis. 
International Journal of Technological Learning, 
Innovation and Development 14:153–191. 
https://doi.org/10.1504/IJTLID.2022.121479 
[25] Bhutani M, Lall B, Dixit A (2021) Mac layer 
performance modelling for ieee 802.15. 7 based 
on discrete‐time Markov chain. IET 
Communications 15:1883–1896. 
https://doi.org/10.1049/cmu2.12199 
[26] Qi J, Meng H, Ye J (2022) A Research on the 
Selection of Cooperative Enterprises in School-
Enterprise Joint Cryptography Laboratory. In: 
2022 International Conference on Artificial 
Intelligence in Everything (AIE). IEEE, pp 659–
663. Doi: 10.1109/AIE57029.2022.00129