


































Global Research in Higher Education 
ISSN 2576-196X (Print) ISSN 2576-1951 (Online) 

Vol. 5, No. 3, 2022 

www.scholink.org/ojs/index.php/grhe 

19 

 

Original Paper 

Key Influencing Factors Affecting the Student Academic 

Performance and Student Satisfactions Ratings: Evidence from 

Undergraduate Students in China 

Lizhi Sun
1
, Danhui Chen

1*
 & Qingping Yang

2
 

1
 Qufu Normal University, Qufu, Shandong, P. R. China 

2
 Brunel University London, Kingston Lane, Uxbridge, UB8 3PH, UK 

∗ Danhui Chen, Qufu Normal University, Qufu, Shandong 273165, P. R. China  

 

Received: July 12, 2022        Accepted: July 29, 2022      Online Published: September 5, 2022 

doi:10.22158/grhe.v5n3p19               URL: http://dx.doi.org/10.22158/grhe.v5n3p19 

 

Abstract   

This paper has developed a sound and practical method to evaluate the key teaching quality including 

the student academic performance and student satisfaction ratings. The method makes use of the 

existing data already readily available in a Chinese university, focusing on the identification of key 

influencing factors affecting the student academic performance and student satisfactions ratings. The 

data analyses have shown the university student academic performance is significantly affected student 

gender, age, previous academic performance, settlements and occupations of parents. There is 

significant difference in the student ratings for different genders and academic positions of teaching 

staff. The student performance and satisfaction ratings also significantly vary in different years of 

intakes and different Schools/programs. The student’s university academic performance can be 

accurately predicted using artificial neural networks with a prediction error of about 7%. This 

approach can help the university to improve the student academic performance and student 

satisfactions. 

Keywords   

higher education quality, internal quality evaluation, data analysis, hypothesis testing, ANOVA, 

artificial neural network 

 

 

 

 



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

20 
Published by SCHOLINK INC. 

 

1. Introduction 

Chinese higher education sector has experienced a series of reforms within last three decades, including 

decentralisation, introduction of market incentives, university merger, internationalisation and 

expansion of higher education (Li, 2010). The continuing reform of Chinese higher education has been 

largely driven by its national policy of domestic social-economic reform and open door policy over last 

three decades. In line with the worldwide trends, quality has become increasingly important in Chinese 

higher education. In particular, the extensive expansion of China’s higher education since late 1990s 

has raised many concerns about the quality of Chinese higher education. Chinese higher education has 

now entered a stage of connotative development of quality improvement. The National Medium-and 

Long-Term Education Reform and Development Plan (2010-2020) (Chinese Ministry of Education, 

2010) points out that improving quality is the core task of higher education development and the basic 

requirement for building a country with strong higher education. The Party leadership's proposals for 

formulating the 14th Five-Year Plan (2021-2025) for National Economic and Social Development and 

the Long-Range Objectives through the Year 2035 was adopted in October 2020. The document has 

clearly put forward the construction of a high-quality education system, which has become the main 

topic of China’s education development (Zhang, 2022).  

In order to improve the quality of higher education, it is necessary to evaluate higher education quality 

and identify the factors affecting the quality. However, there are different opinions about quality in the 

context of higher education. In this paper, the concept of quality is based on the ISO 9000 standard, 

defined as the “degree to which a set of inherent characteristics of an object fulfils requirements” of the 

customers and stakeholders (including students, their future employers and society) (ISO, 2015). The 

student academic performance clearly represents the requirements of students’ future employers, 

whereas the student ratings of the teaching reflect the student satisfaction.  

Evaluation of higher education quality may be carried out at various levels (e.g., institution, program or 

module level) and by different parties (either external or internal). Whilst external evaluation is useful 

for external quality assurance, the importance of internal quality evaluation is increasingly recognised 

since it can facilitate the quality control, quality assurance and also quality improvement (Li, 2010).  

Student quality is a fundamental manifestation of the quality of higher education. For example, M. 

Frazer (Chen, 2004a) believes that the quality of higher education refers first to the quality of student 

development. The academic achievement of university/ college students is the most important 

benchmark to compare the quality of college students (Chen, 2004b). It is also the basic criteria for 

university/college to monitor the quality of teaching and learning and for employers to evaluate and 

select the students.    

As the students are customers of higher education, their experience and satisfaction are also important 

part of higher education quality. Whilst the student satisfactions may cover a range of issues, this paper 

only considers the student satisfaction at module (subject) level.  



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

21 
Published by SCHOLINK INC. 

 

This paper will focus on the quality evaluation and quality improvement at program and module levels 

and present an approach based on the existing data already available from a case study Chinese 

university. The quality evaluation will be centred on the student academic performance (i.e., the 

weighted average of all the modules studied over the 4 year duration of the program) and the student 

evaluation of each module. The paper will perform in-depth analyses and discussions of the important 

influencing factors affecting the student academic performance and the student satisfaction ratings, 

including student’s personal, social economic background and educational factors.   

 

2. Methodology 

2.1 Data Sample 

It is possible to collect a lot of data about the students and the teaching in a university, either existing or 

specially collected for the study. This paper has tried to make the best use of the existing data. A total 

of 1,085 undergraduate students (275 female and 810 male) were selected from the student intakes 

from year 2011 to 2013 from three undergraduate programs within three Schools (Education, Math and 

Electronics Information) of the case study University Q. Their examination results over four year 

studies, together with their national university entrance examination results, personal and 

socio-economic background information were collected. The entrance examination results included 

five core subjects: Chinese, English, Math, Comprehensive Science and Basic Ability. The student 

examination performance was determined using the weighted average marks for the entire four year 

duration of their studies.    

Additionally, another set of data about the student ratings of teaching staff are collected from three 

Schools (Education, Math and Physics) and the data cover 1343 modules taught over 4 years from 

2013/14 to 2016/17. 

2.2 Data Analysis 

Several kinds of data analyses are performed, including exploratory, confirmative and modelling. The 

exploratory data analyses mainly include descriptive statistics and correlation analysis, whilst the 

confirmative data analyses are statistical hypothesis testing, e.g., t-test and ANOVA.   

The purpose of these analyses is to identify the main influencing factors affecting the student 

performance and student satisfaction. These understanding will help the university to identify suitable 

ways to enhance the student learning and the quality of its programs and teaching.  

The correlation analysis calculates Pearson correlation coefficients to measure the linear relationships 

among the five university entrance examination subjects, the University Entrance Examination Total 

Marks (UEETM) and the University Weighted Average Marks (UWAM). The correlation analysis can 

show deep insight into the relationships between subjects and overall academic performance. This 

understanding also helps to evaluate and model the academic performance and the correlated variables 

are further used as the input in the ANN modelling.  



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

22 
Published by SCHOLINK INC. 

 

The factors affecting the student performance are based on the available data and can be broadly 

grouped into the following categories: 

Personal: student gender, age and previous academic performance (university entrance examination 

marks, fresh vs former High School Graduate (HSG)); 

Social-economic factors: rural vs urban, high school area and parents’ occupations; 

Educational: year of intake, different university schools/programmes; 

Based on the exploratory analysis of the data and literature review, the following hypotheses have been 

identified in Table 1, concerning the factors affecting the student’s UEETM and UWAM for each 

student over the whole duration of 4 years studies. 

 

Table 1. Hypotheses Concerning the Factors Affecting the Student UEETM and the Student’s 

University UWAM 

Influencing factors Hypotheses 

Personal 

Gender H1A: The UEETM is equal for male and female students. 

H1B: The UWAM is equal for male and female students. 

Age H2A: The UEETM is equal for students at different ages. 

H2B: The UWAM is equal for students at different ages. 

Fresh vs former 

HSG 

H3A: The UEETM is equal for fresh and former HSG students. 

H3B: The UWAM is equal for fresh and former HSG students. 

Social-economic 

Rural vs urban H4A: The UEETM is equal for students from rural and urban areas. 

H4B: The UWAM is equal for students from rural and urban areas. 

High school 

areas 

H5A: The UEETM is equal for students from high schools in different 

areas. 

H5B: The UWAM is equal for students from high schools in different 

areas. 

Parents’ 

occupation 

H6A: The father’s occupations have no significant influence on the 

UEETM. 

H6B: The father’s occupations have no significant influence on the 

UWAM. 

H7A: The mother’s occupations have no significant influence on the 

UEETM. 

H7B: The mother’s occupations have no significant influence on the 

UWAM. 

Educational 

Year of intake H8A: The UEETM is equal for students from different years of intake. 

H8B: The UWAM is equal for students from different years of intake. 

School/program H9A: The UEETM is equal for students from different school/programs. 

H9B: The UWAM is equal for students from different school/programs. 

 



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

23 
Published by SCHOLINK INC. 

 

The student rating of each module and teaching staff is important for the university to evaluate its 

teaching quality from students’ perspective. Table 2 lists the hypotheses to be tested concerning factors 

affecting the student ratings.  

 

Table 2. Hypothesis Concerning the Student Rating of Each Module and Teaching Staff 

Hypothesis 

H10: The mean student ratings are equal for male and female teaching staff. 

H11: The mean student ratings are equal for staff with different academic positions. 

H12: The mean student ratings are equal for different years of intake. 

H13: The mean student ratings are equal for different school/programs. 

Note. These hypotheses in Tables 1 and 2 are tested using either two sample t-test or ANOVA, with a 

significance level of 0.05.   

 

2.3 Data Modelling Using Artificial Neural Networks 

Data modelling has also been used to study the relationships between various influencing factors and 

the student academic performance. Since the model is rather complex with nonlinear relationships, 

traditional statistical models (e.g., linear or nonlinear regression) are unlikely suitable. Artificial Neural 

Networks (ANNs) have been used for modelling very complex relationships. Cybenko (1989) and 

Funahasi (1989) have demonstrated that ANNs can uniformly approximate arbitrary continuous 

mapping, although the proofs were centred on the existence of such ANNs. The ANNs have found 

successful applications in signal processing, quality engineering and education (Yang, 1994; Kardan, 

2013; Bahadir, 2016; Hu, 2017).  

The architecture of ANN which was modelled earlier by Authors (2018) was adopted in this study and 

it is characterised with one input layer, two hidden layers and one output layer. The detailed design and 

performance studies can be found in (Authors, 2018). The input layer has 11 input variables, including 

student gender, location, fresh or former HSG, high school area, parents’ occupations, and the entrance 

exam results including Chinese, English, Maths, Comprehensive Science and Basic Ability. The output 

is the student academic performance CGPA, based on the standardized Cumulative Grade Point 

Average (CGPA) for the entire four year duration of their studies. Similar to (Authors, 2018), 

Levenberg-Marquardt algorithm was used as the backpropagation training rule. The ANN performance 

is further evaluated through ANN prediction errors. 

 

 

 

 

 



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

24 
Published by SCHOLINK INC. 

 

3. Results and Discussions 

3.1 Correlation Analysis 

The correlation analysis of university entrance exam subject marks, UEETM and UWAM is shown in 

Table 3.   

 

Table 3. Correlation Analysis of University Entrance Exam Subject Marks, UEETM and UWAM 

 Chinese Math English 

Comprehensive 

Science 

Basic 

Ability UEETM UWAM 

Chinese 1       

Math -0.186 1      

English 0.140 0.070 1     

Comprehensive 

Science -0.265 0.031 -0.281 1    

Basic Ability 0.099 0.101 -0.140 -0.011 1   

UEETM 0.160 0.441 0.346 0.540 0.082 1  

UWAM 0.111 0.012 0.255 0.068 0.142 0.230 1 

 

Table 3 shows the UEETM has a strong correlation with English, Math and Comprehensive Science, 

whereas the UWAM has strong correlation with English and UEETM. It is interesting to note that 

Comprehensive Science has a strong negative correlation with Chinese and English, also between Math 

and Chinese. This is likely due to the conflicts in student’s interests and/or study time allocations. It is 

also important to note Math and Comprehensive Science have very weak correlation with the UWAM. 

3.2 Testing of Hypothesis Concerning the Influencing Factors on the Student Academic Performance 

3.2.1 Student Gender’s Influences on Academic Performance 

The t-test results of both the UEETM and the UWAM based on students’ gender are shown in Table 4.   

 

Table 4. Hypothesis Testing Concerning the Influence of Student Gender on UEETM (H1A) and 

UWAM (H1B) 

Hypothesis H1A H1B 

Variable Female Male Female Male 

Mean 547.1 538.4 81.1 75.5 

Variance 321.9 318.9 17.4 20.8 

Observations 810 275 810 275 

df 1083 1083 

t Stat 6.98 18.82 

p-value 0.000 0.000 



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

25 
Published by SCHOLINK INC. 

 

In Table 4, the p-values in both cases are far less than 0.05, indicating a significant difference in both 

the UEETM and the UWAM performance of male and female students, where female students score 

better than male students. Some researchers believe that this is due to that girls’ non-intellectual factors 

are better than boys’, girls have stronger desires and motivations to get better grades than boys, thus 

investing more time and working harder to learn (Wang, 2005). Some studies also believe that the 

current university teaching model is still based on the teacher’s step-by-step systematic teaching, and 

the use of students’ participation and inquiry-based learning methods is not high. The examinations 

also tend to focus on the memory of system knowledge, requiring little creative application of 

knowledge (Ge, 2005). There are also obvious differences in the way of learning between male and 

female students. Girls are better than boys in class notes and recall, whilst boys are better than girls in 

practical application of knowledge. Therefore, female students often tend to achieve better academic 

achievement.  

3.2.2 Student Age’s Influences on Academic Performance 

The t-test results of both the UEETM and the UWAM based on students’ age are shown in Table 5. 

The results show a significant difference between the UWAM for two age groups of students, where 

younger student group score better than older student group. However, the differences in the UEETM 

are not significant. 

 

Table 5. Hypothesis Testing Concerning the Influence of Student Age on the UEETM (H2A) and 

the UWAM (H2B) 

Hypothesis H2A H2B 

Variable Age L Age H Age L Age H 

Mean 544.0 545.8 80.3 79.1 

Variance 372.9 295.8 23.8 23.9 

Observations 542 542 542 542 

df 1082 1082 

t Stat -1.56 4.07 

p-value 0.119 0.000 

 

3.2.3 Influence of Fresh or Former HSG on Academic Performance 

Table 6 shows the t-test results of the differences in both UEETM and UWAM based on whether or not 

the student is a former or fresh High School Graduate (HSG). Whilst the differences in the UEETM are 

not significant, there is a significant difference in the UWAM for the two student cohorts, where fresh 

HSG students score significantly higher than former HSG students. A study by Fu et al. (2022) has 

confirmed this point of view: the Grade Point Average (GPA) of different types of students from 



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

26 
Published by SCHOLINK INC. 

 

highest to lowest are: rural fresh HSG students, urban fresh HSG students, urban former HSG students, 

and rural former HSG students. 

 

Table 6. Hypothesis Testing Concerning the Influence of Fresh or Former HSG on UEETM 

(H3A) and UWAM (H3B) 

Hypothesis H3A H3B 

Variable Former HSG Fresh HSG Former HSG Fresh HSG 

Mean 543.8 545.4 78.2 80.3 

Variance 277.0 358.8 24.6 22.8 

Observations 315 770 315 770 

p-value 0.214 0.000 

 

3.2.4 Settlement’s Influences on Academic Performance 

The t-test results of the differences in both the UEETM and the UWAM based on rural and urban 

settlements are presented in Table 7. 

 

Table 7. Hypothesis Testing Concerning the Influence of Location on UEETM (H4A) and 

UWAM (H4B) 

Hypothesis H4A H4B 

Variable Rural Urban Rural Urban 

Mean 545.3 543.7 79.8 79.2 

Variance 331.7 346.1 23.3 26.9 

Observations 824 261 824 261 

df 1083 1083 

t Stat 1.18 1.65 

p-value (two tail) 0.237 0.099 

p-value (one tail) 0.119 0.049 

 

As shown in Table 7, the differences in the UEETM of rural and urban areas are not statistically 

significant. But the one-tailed t-test of UWAM has a p-value marginally smaller than 0.05 (two-tailed 

test with p-value of 0.099), indicating the UWAM of the rural students is significantly higher than that 

of the urban students. This result is similar to that reported in (Chu, 2011): the college students with 

higher income levels from cities have lower academic performance than the students from rural areas 

with lower family incomes. Coleman (1966) pointed out that family background is the most important 

factor affecting academic achievement, and its influence runs through the whole process of students’ 

studies, and the influence has an accumulating effect. Under China’s urban-rural dual system, there is a 



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

27 
Published by SCHOLINK INC. 

 

big gap between urban family conditions and rural family conditions. Different levels of capital 

provided to the children by families with different backgrounds have an impact on college students’ 

input to learning (Liu, 2015), which affects academic performance. Generally, college students from 

county towns, townships and rural areas have significantly higher learning input than college students 

from provincial capital cities (Li, 2022). In addition, the psychological quality of students is affected by 

the family environment and conditions. The students with poor economic conditions tend to have more 

positive learning attitudes and stronger learning motivation. Rural college students are more diligent 

and hard-working than urban students, with good psychological qualities such as hard working and 

perseverance. These good learning qualities are an important factor for college students to achieve 

excellent results. 

3.2.5 Student High School Areas’ Influences on Academic Performance 

In order to test if there is a significant difference between the student performances from different high 

school areas, ANOVA has been carried out of both the UEETM and UWAM in 17 different high 

school areas, with the results shown in Tables 8A/8B. The results have shown no significant 

differences in both UEETM and UWAM due to different high school areas. 

 

Table 8A. ANOVA of the UEETM in Different High School Areas (H5A) 

Source of Variation SS df MS F P-value F crit 

Between Groups 4620.125 16 288.7578 0.859455 0.61712 1.652984 

Within Groups 358824.1 1068 335.9776    

       

Total 363444.2 1084     

 

Table 8B. ANOVA of the UWAM in Different High School Areas (H5B) 

Source of Variation SS df MS F P-value F crit 

Between Groups 468.4369 16 29.27731 1.353055 0.157541 1.652984 

Within Groups 23109.3 1068 21.63792    

       

Total 23577.74 1084     

 

3.2.6 Student Parent’s Occupation on Academic Performance 

Table 9 combines the two ANOVA tables for testing the differences in both the UEETM and the 

UWAM due to the different occupations of fathers and mothers. The results have shown the differences 

in the UEETM due to parents are not significant. But both parents have significant influences on the 

UWAM, with mothers (p-value of 0.016) having more significant influences than fathers (p-value 

0.082). There are 6 categories of occupations for a parent: unspecified (0), unemployed (1), farmers (2), 



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

28 
Published by SCHOLINK INC. 

 

self-employed or factory workers (3), professional (teachers or engineers or doctors) (4) and officers or 

managers (5). The differences in the UEETM are not significant for either fathers’ or mothers’ 

occupations, but the differences in the UWAM are significant for mothers’ occupational influence. The 

fathers’ occupational influences on the UWAM are not significant at a significance level of 0.05, but 

significant at a significance level of 0.10. 

 

Table 9. ANOVA of the UEETM and UWAM due to Occupational Influences of Fathers 

(H6A/H6B) and Mothers (H7A/H7B) 

 UEETM UWAM 

Programme df F p-value F p-value 

Father (5,1079) 1.474 0.196 1.963 0.082 

Mother (5,1079) 1.504 0.186 2.808 0.016 

 

We believe that the relationship between parental occupation and student academic performance is 

influenced by the mediating factor of learning involvement (Crook, 1995; Ermisch, 2010; Marks, 2008). 

Nord (1998) found that the higher the father’s level of learning involvement with the child, the better 

the child’s academic performance is. Fathers with lower levels of education are usually less involved in 

student learning. Similarly, the occupation type of Chinese mothers has a great influence on children’s 

academic performance. It can be explained that there is always a tradition of “male outside, female 

inside” in China. In many families, mothers and children have more contact when their children grow 

up. In China, Pang (2013) and Li (2020) and other studies have found that the level of education of 

mothers has a more significant impact on students’ academic performance than their fathers. 

Close examination reveals the differences in the UWAM are significantly caused by category 4 

occupation (professionals including teachers, engineers and doctors), the further ANOVA of the other 5 

categories shows no significant differences in the UWAM. The students with a professional mother 

have performed significantly lower than the other group of students. The students with a professional 

father have also performed lower than the other group of students, although only significant at a 

significant level of 0.10. 

The significant differences are likely due to that professional parents are often very busy and have 

therefore less contact with their children. 

3.2.7 Different Intake’s Influences on Academic Performance 

Table 10 shows the ANOVA results of both the UEETM and the UWAM in different years of intakes. 

The comparison has been made for each of three School/Programs, Education, Mathematics and 

Electronics Information. The ANOVA results have shown that there are significant differences 

(p-value=0.000) in the UEETM for all three School/Programs. However, for the UWAM, only 

Mathematics has significant difference in the performances of three years’ intakes. 



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

29 
Published by SCHOLINK INC. 

 

Table 10. ANOVA of the UEETM (H8A) and the UWAM (H8B) in Different Intakes  

 UEETM UWAM 

School/Program df F p-value F p-value 

Education (2,325) 91.521 0.000 1.092 0.337 

Mathematics (2,315) 102.609 0.000 5.944 0.003 

Electronics Info. (2,436) 559.989 0.000 2.117 0.122 

 

3.2.8 Different School/Program’s Influences on Academic Performance 

Table 11a combines two ANOVA tables to test the differences in both the UEETM and UWAM for 

different School/Programs. The ANOVA results have shown significant differences in both the 

UEETM and the UWAM among the three School/Programs. Table 11b shows that the UEETMs for 

Education and Mathematics programs are significantly higher than the Electronics Information 

Program and that in the UWAM, Education School/Program scores significantly higher than other two 

Schools/Programs. 

 

Table 11. ANOVA of the UEETM (H9A) and the UWAM (H9B) in Different School/Programs 

a) Combined ANOVA tables 

 UEETM UWAM 

ANOVA of 3 

different 

School/Programs 

df F p-value F p-value 

(2,1082) 217.9361207 0.000 84.17819034 0.000 

 

b) Performance statistics for each School/Program 

  UEETM UWAM 

Groups Count Average Variance Average Variance 

Edu 328 552.64 194.74 82.03 15.02 

Math 318 553.37 220.14 79.94 23.84 

Electronics 439 533.02 286.78 77.72 23.31 

 

3.3 Testing of Hypotheses Concerning the Influencing Factors on Student Satisfaction Ratings  

3.3.1 Teacher Gender’s Influences on the Student Ratings 

In Table 12, the two-sample t-Test performed has shown the mean ratings for male teaching staff are 

significantly higher than female teaching staff, but the significance varies amongst the three Schools. 

The difference in the ratings of male and female staff is only significant (t(206)=-3.35, p-value=0.000) 

in Math School/Program, but not in Education (t(331)=-0.75, p-value=0.452) and Physics (t(333)=-1.36, 

p-value=0.175).    



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

30 
Published by SCHOLINK INC. 

 

Table 12. T-test of the Mean Student Ratings of the Lectures for Different Gender of Teaching 

Staff 

Variable Female Male 

Mean 94.84 95.04 

Variance 1.630 1.069 

Observations 493 850 

df 865 

t Stat -2.95 

p-value 0.003 

 

3.3.2 Teacher Academic Position’s Influences on the Student Ratings 

Hypothesis H11 is used to test if the mean average of the student ratings are equal for different 

academic positions. In the Education School, a two-sample t-Test indicates that the ratings for 

Teaching Assistants are significantly (t(13)=4.43, p-value=0.000) lower than Lecturers/Associate 

Professors/Professors. But the single factor ANOVA shows that the ratings are not significantly (F(2, 

464)=0.41, p-value= 0.661) different amongst the Lecturers, Associate Professors and Professors. 

In the Math School, there are significant (F(2,375)=4.13, p-value= 0.017) difference in the ratings of 

the Lecturers, Associate Professors and Professors, with the mean ratings of 94.9, 95.1 and 95.3, 

respectively. In Physics School, the evaluated staff also include teaching Technicians and Senior 

Technicians. The ANOVA has shown that there are significant (F(4,473)=4.13, p-value= 0.017) 

differences amongst the ratings of the five categories of teaching staff, with the decreasing rating 

means of 95.5, 95.2, 95.0, 94.8 and 94.7 for Professors, Technicians, Associate Professors, Senior 

Technicians and Lecturers, respectively. 

3.3.3 Different Intake’s Influences on the Student Ratings 

Hypothesis H12 is used to test if the mean average of the student ratings of the lectures are equal for 

different intakes. The ANOVA has indicated that there are significant (F(3,1339)= 28.5, P-value=0.000) 

difference in the ratings of 4 year intakes, with the mean ratings of 94.84, 94.53, 95.05 and 95.41 for 

2013/14, 2014/15. 2015/16 and 2016/17, respectively.  

3.3.4 Different School’s Influences on the Student Ratings 

Table shows the ANOVA table for the testing of different school’s influences on the student ratings. 

The ANOVA has shown that there are significant (F(2,1340)= 4.29, P-value=0.014) difference in the 

ratings of 3 Schools, with the mean ratings of 94.85, 95.05 and 95.03 for Education, Math and Physics, 

respectively. 

 

 



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

31 
Published by SCHOLINK INC. 

 

Table 13. ANOVA Table for the Testing of the Mean Student Ratings for the Three Different 

Schools 

Source of Variation SS df MS F P-value F crit 

Between Groups 10.95459 2 5.477295 4.289398 0.013902 3.00244 

Within Groups 1711.097 1340 1.276938    

       

Total 1722.052 1342     

 

According to the two-sample t-Test performed, the ratings for the Math/Physics Schools are 

significantly (t(842)=-2.76, p-value=0.006) higher than the Education School, but the differences 

between Math and Physics Schools are insignificant (t(845)=0.25, p-value=0.806). 

3.4 ANN Modelling  

Based on the ANN modelling approach by Authors (2018), the ANN model of the university academic 

performance is performed using Matlab software. The input layer consists of 11 variables about 

students’ personal/social economic background (i.e., gender, location, fresh or former HSG, high 

school area, parents’ occupations) and the entrance exam results including Chinese, English, Maths, 

Comprehensive Science and Basic Ability. Each of the two hidden layers has 30 hidden neurons and 

the single output neuron predicts the students’ CGPA.  

 

 

Figure 1. ANN Prediction Error Histogram  

 

 

 

 

 



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

32 
Published by SCHOLINK INC. 

 

Data samples are randomly mixed and divided into three subsets for training, validation and test with a 

ratio of 0.7:0.15:0.15, respectively. Levenberg-Marquardt algorithm is used during the training and 

learning with a damping factor ζ of 0.001 and the training epoch of 1,500. The ANN training continues 

until the validation error failed to decrease for six iterations during the validation process. After 

training and validation runs, the ANN has achieve a performance with the MSE≈0.27 or 6.9%, 

demonstrating a good prediction performance, with the error histogram shown in Figure 1. 

 

Table 14. Summary of the Testing Results of the Hypotheses Concerning the Factors Affecting 

the Student Academic Performance 

Influencing factors Hypotheses Decision 

Personal Gender H1A: The UEETM is equal for male and female students. Reject 

H1B: The UWAM is equal for male and female students. Reject 

Age H2A: The UEETM is equal for students at different ages. Accept 

H2B: The UWAM is equal for students at different ages. Reject 

Fresh vs former 

HSG 

 

 

H3A: The UEETM is equal for fresh and former HSG students. Accept 

H3B: The UWAM is equal for fresh and former HSG students. Reject 

Social-economic Rural vs urban H4A: The UEETM is equal for students from rural and urban areas. Accept 

H4B: The UWAM is equal for students from rural and urban areas. Reject 

High school areas H5A: The UEETM is equal for students from high schools in 

different areas. 

Accept 

H5B: The UWAM is equal for students from high schools in 

different areas. 

Accept 

 

Parents’ 

occupation 

H6A: The father’s occupations have no significant influence on the 

UEETM. 

Accept 

H6B: The father’s occupations have no significant influence on the 

UWAM. 

Reject 

@α=0.10 

H7A: The mother’s occupations have no significant influence on the 

UEETM. 

Accept 

H7B: The mother’s occupations have no significant influence on the 

UWAM. 

 

Reject 

Educational Year of intake H8A: The UEETM is equal for students from different years of 

intake. 

Reject 

H8B: The UWAM is equal for students from different years of 

intake. 

Note
1
 

School/program H9A: The UEETM is equal for students from different 

school/programs. 

Reject 

H9B: The UWAM is equal for students from different 

school/programs. 

Reject 

Note. Reject for Mathematics Program, Accept for Education and Electronics Information Programs. 

 



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

33 
Published by SCHOLINK INC. 

 

Table 15. Summary of the Testing Results of the Hypotheses Concerning the Student Rating of 

Each Module and Teaching Staff 

Hypothesis Decisio

n 

H10: The mean student ratings are equal for male and female teaching staff. Reject 

H11: The mean student ratings are equal for staff with different academic positions. Reject 

H12: The mean student ratings are equal for different years of intake. Reject 

H13: The mean student ratings are equal for different school/programs. Reject 

 

4. Conclusion 

This paper has developed a practical approach to evaluate the quality of university teaching and 

learning using the existing data which are already readily available. It aims to identify the important 

influencing factors affecting the student academic performance and the student satisfaction rating of 

teaching staff and modules. The study has considered student’s personal, social economic background 

and educational factors. The relationship between the key factors and student academic performance 

has been further modelled with Artificial Neural Networks (ANNs).   

The findings of the data analyses and hypothesis testing are summarised in Tables 14 and 15, which 

present the overall patterns of the relationships between the various influencing factors and the learning 

and teaching quality in terms of academic performance and student satisfaction ratings. The following 

can be concluded from these results: 

1) The university student academic performance is significantly affected student gender, age, previous 

academic performance, settlements and occupations of parents (particularly mother’s occupation).  

2) Male, older, urban, former HSG students and those with professional parent(s) tend to have 

significantly lower performance than their counterparts.   

3) The student academic performance significantly varies in different years of intakes and different 

Schools/programs. 

4) Based on the case study of University Q, there is significant difference in the student ratings for 

different genders and academic positions of teaching staff. 

5) The student satisfaction ratings also significantly vary in different years of intakes and different 

Schools/programs. 

6) The student’s university academic performance can be quite accurately predicted using ANN with a 

prediction error of about 7%. 

These findings are based on the limited data samples from three schools of the case study university. 

Similar studies (including data analyses and modelling) can be extended to more schools and more 

universities. These types of findings can be used to evaluate, diagnose and predict the quality of the 

learning and teaching. They will also help the university to identify suitable practices and resources to 



www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

34 
Published by SCHOLINK INC. 

 

enhance and improve the student academic performance and student satisfactions, contributing to the 

continual improvement of higher education quality. 

 

Funding 

This study was part of the education project “Research on Teaching Investment of Chinese University 

Faculty and Improvement of Undergraduate Teaching Quality” (Project Number: BIA200209), funded 

by the National Social Science Fund of China. 

 

References 

Bahadir, E. (2016). Using neural network and logistic regression analysis to predict prospective 

mathematics teachers’ academic success upon entering graduate education. Educational Sciences: 

Theory & Practice, 16(3), 943-964.  

Chen, Y. (2004a). Introduction to higher education quality assurance system. Beijing Normal 

University Press. 

Chen, Y. et al. (2004b). Introduction to the quality assurance system in higher education. Beijing 

Normal University Press. 

Chinese Ministry of Education. (2010). Outline of China’s national plan for medium-and long-term 

education reform and development (2010-2020). 

Chu, Y. (2011). Research on the correlation between household register and college student academic 

performance-Empirical analysis of three universities in Guangdong Province. Higher Education 

Exploration, 3, 110-116. 

Coleman, J. et al. (1966). Equality of education opportunity. In Report No OE-138001. National Centre 

for Educational Statistics, Washington DC. 

Crook, C. J. (1995). The role of mothers in the educational and status attainment of Australian men and 

women. Zeal Australian and New Journal of Sociology, 31(2), 45-73. 

https://doi.org/10.1177/144078339503100202 

Cybenko, G. (1989). Approximation by superpositions of a sigmoidal function. Mathematical Control 

Signals Systems, 2, 303-314. https://doi.org/10.1007/BF02551274 

Ermisch, J., & Pronzato, C. (2010). Causal effects of parents’ education on children’s education. In 

ISER Working Paper Series, No 2010-16. Institute for Social and Economic Research.  

Fu, X. Z. et al. (2022). Individual Influencing Factors of College Students’ Academic Achievements 

and Their Mechanisms. University Education, 4, 251-254. 

Funahashi, K. (1989). On the approximate realization of continuous mappings by neural networks. 

Neural Networks, 2, 183-192. https://doi.org/10.1016/0893-6080(89)90003-8 

Ge, M., & Yu, Y. (2005). College students’ learning adaptability and its influencing factors. Journal od 

of Anhui Normal University (Humanities and Social Sciences Edition), 9, 602-606. 

https://doi.org/10.1177/144078339503100202
https://doi.org/10.1007/BF02551274
https://doi.org/10.1016/0893-6080(89)90003-8


www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

35 
Published by SCHOLINK INC. 

 

Hu, B. (2017). Teaching quality evaluation research based on neural network for university physical 

education. In 2017 International Conference on Smart Grid and Electrical Automation (ICSGEA) 

(pp. 290-293). https://doi.org/10.1109/ICSGEA.2017.155 

ISO. (2015). ISO 9000:2015, Quality management systems fundamentals and vocabulary. 

Kardan, A. et al. (2013). Prediction of student course selection in online higher education institutes 

using neural network. Computers & Education, 65, 1-11. 

https://doi.org/10.1016/j.compedu.2013.01.015 

Lau, E. T., Sun, L., & Yang, Q. (2019). Modelling, prediction and classification of student academic 

performance using artificial neural networks. SN Applied Sciences, 1(9), 1-10. 

https://doi.org/10.1007/s42452-019-0884-7 

Li, L. L. (2022). The Empirical Study on the Relationship between College Student Engagement and 

the Cultural Capital in Western China. Higher Education of Sciences (To be submitted). Retrieved 

from https://kns.cnk i.net/kcms/detail/62.1028.g4.20220531.1459.002.html 

Li, Y. (2010). Quality assurance in Chinese higher education. Research in Comparative and 

International Education, 5(1), 58-76. https://doi.org/10.2304/rcie.2010.5.1.58 

Li. X. J. (2020). Empirical Analysis of the Influence of Family Educated Levels Capital on College 

Student Engagement. Shanghai Journal of Educational Evaluation, 1, 64-69. 

Liu, P. (2015). Research on College Students’ learning adaptability (PhD Thesis). Shanghai Normal 

University. 

Liu, X. (2015). Differences in learning input among college students with different family backgrounds 

(Master Dissertation). Fujian Normal University. 

Marks, G. N. (2008). Are father’s or mother’s socioeconomic characteristics more important influences 

on student performance? Recent international evidence. Social Indicators Research, 85(2), 

293-309. https://doi.org/10.1007/s11205-007-9132-4 

Nord, C. W. (1998). Students do better when their fathers are involved at school. In Report No 

NCES-1998-121. National Centre for Educational Statistics, Washington DC. 

Pang, W. et al. (2013). A study on the relationship between family socioeconomic status and academic 

achievement of middle school students. Global Education, 42(2), 12-20. 

Wang, A., & Che, H. (2005). A study on the relationship between learning anxiety, learning attitude 

and involvement motivation and academic achievement-A survey of learning experience of 

Psychological Statistics. Psychological Development and Education, 1, 55-59. 

Yang, Q., & Butler, C. (1994). Sensor signal processing using neural networks for a 3-D fibre-optic 

position sensor. Sensors and Actuators, A: Physical, 41(3), 102-109. 

https://doi.org/10.1016/0924-4247(94)80096-0 

https://doi.org/10.1109/ICSGEA.2017.155
https://doi.org/10.1016/j.compedu.2013.01.015
https://doi.org/10.1007/s42452-019-0884-7
https://doi.org/10.2304/rcie.2010.5.1.58
https://doi.org/10.1007/s11205-007-9132-4
https://doi.org/10.1016/0924-4247(94)80096-0


www.scholink.org/ojs/index.php/grhe            Global Research in Higher Education                  Vol. 5, No. 3, 2022 

36 
Published by SCHOLINK INC. 

 

Zhang, J. M. (2022). The construction of intercollegiate relations of cooperative colleges and universes 

in the context of high-quality high education system system. Journal of Higher Education 

Management, 16(4), 20-30. 

 


