Research on the Influencing Factors of College Students' Academic

Achievement Based on the Correlation Test Approach

Liuqi Chen

School of Cyber Science and Engineering, Nanjing University of Science and Technology, Jiangsu, China

Keywords: Academic Achievement, Correlation Test, Gender, Age of Enrollment, Multiclassification Logistic

Regression.

Abstract: College students' academic performance is an important indicator reflecting the quality of college education.

Based on the data provided by the academic system of colleges and universities, this paper uses correlation

test, chi-square test, multiclassification logistic regression analysis and other methods to investigate the effects

of gender, age at enrollment, attendance time, GDP, types of courses and other factors on students' academic

performance. It is found that gender, age at enrollment, attendance time, GDP, and types of courses had some

correlation with academic performance. In particular, students of male gender, older age at school, evening

attendance, and students from economically disadvantaged areas tend to be at a disadvantage in terms of

academic performance. And colleges and universities should pay attention to teaching these students and

guiding them to study efficiently in order to weaken the negative impact of these factors on students. In

addition to this, such students themselves should work harder to avoid academic warnings or even dropping

out of school.

1 INTRODUCTION

With the rapid development of the times, more and

more people are able to receive higher education and

become a college student. The academic performance

of college students is a reflection of the quality of the

education they receive. Unfortunately, the academic

performance of college students is partly polarized.

Many college students have excellent academic

performance, while many others face academic

warnings or even dropout. The study of the factors

affecting the academic performance of college

students is of great significance for improving the

quality of higher education and cultivating better

talents.

Scholars at both domestic and foreign countries

have made a lot of achievements on this research.

However, the previous research on the factors

affecting college students' academic performance is

often limited to a single aspect. In view of this, on the

one hand, this paper will summarize the results of

previous researchers, and on the other hand, it will also

process the data set through correlation test,

multiclassification Logistics regression analysis and

other methods as a supplement to the content of

previous researches.

Li and Omeng argued that the academic

performance of poor students is related to the poverty

level of their families (Li and Ou 2018). They

demonstrated that for poor families, family income

has a positive effect on students' academic

performance. Song used questionnaires and

interviews to illustrate the link between academic

performance and students' space preferences in the

classroom (Song 2020). He found that college

students' classroom spatial preference exists in

educationally active zones, and students in the front

and left-centre zones tend to be more likely to be paid

attention to by the teacher and to communicate with

them more, and therefore will receive more

educational resources and achieve higher academic

performance.

Some scholars have used the propensity score

matching method and the double difference method

based on propensity score matching to verify that

serving as a student leader can have a significant

positive impact on academic achievement (Wu and

Cui 2019). In this study, scholars pointed out that

student leaders tend to be motivated to learn and more

goal-oriented, and these factors all play a role in their

success in academic performance.

Scholars such as Wang focused on exploring the

influence of traditional values on students' academic

Chen, L.

Research on the Inﬂuencing Factors of College Students’ Academic Achievement Based on the Correlation Test Approach.

DOI: 10.5220/0012816600003885

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Data Analysis and Machine Learning (DAML 2023), pages 14-19

ISBN: 978-989-758-705-4

performance (Wang et al 2019). Surprisingly,

although traditional values did not have a significant

effect on public class performance, they had a very

significant positive effect on professional class

performance. Also they pointed out that classroom

silent behaviour is likely to be a medium through

which traditional values act on academic performance.

This is something that needs to be adjusted

accordingly by the teaching staff. Also deserving of

some changes on the part of the teachers are the

teaching styles. A questionnaire experiment has

shown that microteaching has a significant positive

effect on students' motivation and academic

performance (Zhang et al 2020). In addition to this,

teachers need to improve their credibility with the

student body. This has also been shown to have a

positive effect on college students' academic

performance (Li et al 2020).

Hao and other scholars suggested that there might

be a link between the academic performance of

college students and their social circle (Hao et al

2020). They used standard deviation of grades and

spatial autoregressive model to analyze the effect of

social circle on academic performance. The results

shown that a good learning atmosphere within the

social circle positively affects college students'

academic performance.

It is worth mentioning that part-time job or not is

also an important factor affecting students' academic

performance (Xiong and Niu 2022). Xiong Yanqing

and other scholars found that for the vast majority of

the student population, there is a weak conflict effect

between part-time jobs and academic performance

before. In addition, the specific type and duration of

part-time jobs are also related to academic

performance. Meanwhile, there are many factors,

including psychological capital (Song et al 2021),

class atmosphere (Huang and Chen 2021). They also

play a role in the academic performance of college

students. Foreign scholars pay more attention to the

effect of sleep duration on academic performance

(Allison and Michael 2022). They believed that

students who get enough sleep tend to encounter fewer

academic obstacles, which makes it easier for them to

achieve higher academic performance.

Previous studies have focused on a variety of

factors, including social and economic factors.

However, there are still many factors that have not

been considered. In this paper, we will use correlation

tests, multiclassification logistic regression analyses,

and other methods to explore the effects of these

factors on college students' academic performance

from age at enrollment, time of attendance, gender,

and GDP.

2 METHODS

2.1 Data Sources and Description

The data used in this article comes from the Kaggle. It

includes demographic data, social-economic factors

and academic performance information that can be

used to analyze the possible predictors of student

dropout and academic success.

2.2 Selection and Description of

Indicators

Given that previous articles have not explored the

effects of students' age at enrollment, course types,

gender, GDP, and attendance time on students'

academic performance, this paper will examine these

five factors (Table 1).

Table 1: Statistical indicators data table.

variant data type range abbreviation

Age at

enrollment

number 18-35 x1

Daytime/evening

attendance

number 0,1 x2

Gender number 0,1 x3

GDP number -7.76 x4

Semester course

grades

number 0-17.58 x5

Target word

Dropout,

Graduate,

Enrolle

Semester course

assessment

number 0-21 x7

Course number 1-17 x8

2.3 Methodology

2.3.1 T-Test

Often t-tests are used to validate hypothesis tests for

individual normal overall parameters when the

overall variance of the sample of data used for

statistics is unknown. For the variables semester

course assessment and semester course grades, it can

be assumed that their totals conform to a normal

distribution and that the overall standard deviation is

unknown. Therefore, in this paper the t-test will be

used to discuss whether the differences in their means

are significant or not.

Research on the Inﬂuencing Factors of College Students’ Academic Achievement Based on the Correlation Test Approach

2.3.2 Correlation Test

In this paper, Pearson correlation coefficient,

Spearman correlation coefficient will be used to

investigate the correlation between independent and

dependent variables.Spearman correlation coefficient

is a non-parametric measure of the dependence of two

variables on each other. It describes the correlation

between two variables in terms of degree of

monotonicity.Pearson correlation coefficient is used

to assess the linear correlation between two variables.

2.3.3 Chi-Square Test

The chi-square test counts the degree of deviation

between the actual observed values of a sample and

the theoretically inferred values.

2.3.4 Multiclassification Logistic Regression

Analysis

The outcome of whether or not a student drops out of

school is a categorical variable in this paper, as well

as the large number of socio-economic factors

involved, it was decided to use a multiclassification

logistic regression model to investigate the extent to

which these factors influence the outcome of students

dropping out of school.

3 RESULTS AND DISCUSSION

In the processing of the data set, this paper has

produced a heat map for the correlation coefficients of

the indicators. The data in this figure 1 will be used to

provide a discussion of the correlation of the factors

with semester course grades.

Figure 1: Heat map of correlation coefficients between

indicators.

3.1

Age at Enrollment

From Figure 1 this paper found that the Pearson

correlation coefficient between age of enrolment and

semester course grades reached -0.157 and showed

significance at 0.01 level. Thus there is a significant

negative correlation between semester course grades

and the age of students' enrolment. In this regard, this

paper hypothesises that the possible reason for this

may be that younger students are relatively more

smart, while older students may be distracted from

their studies due to many life factors. In order to get

better grades, older students should put in more efforts

in their studies to compensate for the disadvantage of

age.

3.2

Gender

In the dataset, 0 was used to refer to females and 1 was

used to refer to males. In order to investigate the

correlation between gender and semester course

grades, the t-test was used for the study. From Figure

1, it can be seen that all of the samples of different

genders showed significance (p<0.05) for all of the

semester course grades, meaning that all of the

samples of different genders were different for the

semester courses (showing significance at the 0.01

level). And from Figure 2, it can be concluded that the

mean value of 11.32 for females is significantly higher

than the mean value of 9.4 for males.

Figure 2: Comparison of t-test for gender and semester

course grades.

The results show that women perform

significantly better than men in terms of academic

performance at university. In addition to academic

performance, the dropout rate is also a very important

indicator. In this regard, this paper chose to use cross-

tabulation analysis (chi-square test) to explore the

effect of gender on whether or not a student drops out

of school.

x1 x2 x3 x4 x5 x6

1 -0.462 0.149 -0.065 -0.157 0.243

-0.462 1 -0.012 0.023 0.064 -0.075

0.149 -0.012 1 -0.008 -0.189 0.229

-0.065 0.023 -0.008 1 0.055 -0.044

-0.157 0.064 -0.189 0.055 1 -0.485

0.243 -0.075 0.229 -0.044 -0.485 1

DAML 2023 - International Conference on Data Analysis and Machine Learning

Figure 3: Cross-tabulation of gender and dropout outcomes.

Figure 3 shows that 45.05 per cent of males drop

out of school, which is significantly higher than the

25.10 per cent of females who do so. Whereas, 57.91

per cent of females successfully graduated, which is

significantly higher than 35.22 per cent of males. In

this chi-square test the p-value is 0.000<0.05, and the

different gender samples show significant difference

for the students' dropout rate.

In order to exclude the possible existence of the

factor that the teacher's awarding of marks is not fair

to males and females, this paper also did a correlation

test of Pearson coefficient between the semester

course assessment and gender. The results are shown

in Table 2.

Table 2: Table of Pearson's correlation coefficients between

gender and semester course assessment.

Semester course

assessment

correlation coefficient -0.021

gender p-value 0.156

sam

le size 4424

The value of the correlation coefficient between

semester course assessment and gender is -0.021,

which is close to 0, and the p-value is 0.156 > 0.05,

thus indicating that there is no correlation between

semester course assessment and gender. At this point,

the possible factor of unfairness in the teacher's

grading can be excluded, and it is basically certain that

gender has a real impact on the students' learning

status.

University study often requires a great deal of self-

discipline and time management skills. This paper

hypothesises that it is likely that women's academic

performance at university is significantly better than

that of men because of their significant advantages

over men in these two areas. In order to achieve the

desired results in university and avoid failing courses

and academic warnings, males should focus on self-

discipline and time management, and should not lose

their minds in fun and games. Of course, there may be

other factors, and it is up to future researchers to find

out the exact reasons.

3.3

GDP

For the effect of GDP, the Pearson and Spearman

coefficients were used to explore the correlation

between the two variables.

From Figure 1, it can be easily seen that the value

of the correlation coefficient between semester course

grades and GDP is 0.055 and shows significance at

0.01 level. And when this paper starts from the point

of view of Spearman coefficient, this paper gets the

same answer.

Table 3: Spearman's correlation coefficient table between

GDP and semester course grades.

Semester course grades

correlation coefficient 0.092

GDP p-value 0.000

sample size 4424

The value of the correlation coefficient between

semester course grades and GDP is 0.092 and shows

significance at the 0.01 level, thus indicating that there

is a significant positive correlation between semester

course grades and GDP. This author speculates that

this may be due to the higher teaching standards of

teachers and the relatively better supporting facilities

in developed areas. Students are more likely to achieve

satisfactory results.

3.4

Time of Attendance

In addition to this, students' attendance time may also

be an important factor affecting students' academic

performance. In this regard, this paper used t-test to

investigate the effect (figure 4).

Figure 4: Comparison of t-test of attendance times and

semester course grades.

Research on the Inﬂuencing Factors of College Students’ Academic Achievement Based on the Correlation Test Approach

From Table 4, it can be seen that: the different

attendance time samples show significance (p<0.05)

for all of the semester course grades. The mean value

of 10.75 for daytime attendance is significantly higher

than the mean value of 9.76 for evening attendance,

and it is hypothesised that this may be due to the fact

that during the daytime it is easier for students to

concentrate and attend classes more efficiently, and

therefore it is easier for them to achieve higher grades.

In the evening, however, students are more sleepy and

less efficient in class attendance, resulting in

insufficient knowledge mastery and lower test scores.

Considering the above, this article suggests that

students try to choose the daytime courses, and if they

choose the evening courses, they should be more

energetic to listen to the lectures, so as to avoid

achieving lower scores due to lack of concentration.

Table 4: Table of t-test results of attendance hours and

semester course grades.

Attendance time (mean

± standard deviation

)

t p

0.0(n=483) 1.0(n=3941)

Semester

course

grades

9.76±4.94 10.75±4.82 -4.181 0.000**

* p<0.05 ** p<0.01

3.5

Types of Courses

On top of these factors, this paper also considers the

fact that different course difficulties may create

different barriers to graduation difficulty for students.

So the paper makes a cross-tabulation graph of course

type and student dropout outcomes.

Figure 5: Cross-tabulation of programme and dropout

outcomes.

From the figure 5, it can be seen that there are

obvious differences in the failure rates of different

courses. For courses with high failure rates, students

should study carefully to overcome the difficulties,

and teachers should consider lowering the difficulty of

the course assessment to avoid making it too difficult

for students.

3.6

Regression Modelling

Above this paper analyzed the correlation between

each factor and students' academic performance, and

below this paper will use the multiclassification

logistic regression analysis model to describe the

extent of the impact of these factors on students'

academic performance..

Table 5: Summary of the results of the multiclassification

logistic regression analysis.

Enrolled in

school

(regression

coefficient)

Dropout

(regression

coefficient)

gender

0.630** 0.965**

-7.113 -12.926

age at enrollment

0.013 0.072**

-1.662 -12.678

daytime/evening

attendance

0.089 0.265*

-0.546 -2.061

GDP

-0.005 -0.032*

(-0.293) (-2.037)

intercept

-1.585** -2.719**

(-5.721) (-12.648)

likelihood ratio test χ2(8)=466.443,p=0.000

dependent variable: outcome

* p<0.05 ** p<0.01 z-values in brackets

After regression analyses (Table 5), the paper

ended up with two final formulas: ln(enrolled

/graduated) = -1.585 + 0.63 gender + 0.013 age at

school enrollment + 0.089 attendance time - 0.005

GDP. and ln(dropped out / graduated) = -2.719 +

0.965 gender + 0.072 age at school enrollment + 0.265

attendance time - 0.032*GDP.

Relative to graduation, the regression coefficient

value for gender is 0.630 for enrollment in school and

shows significance at the 0.01 level (z=7.113,

p=0.000<0.01), implying that gender significantly and

positively influences the outcome. While the

remaining four indicators are not for presenting

significance on the results. This means that these four

indicators do not have a significant effect on whether

students end up dropping out of school or not.

With this model, we can make predictions about

the likelihood of students successfully graduating.

DAML 2023 - International Conference on Data Analysis and Machine Learning

One of the predictions for graduation had an accuracy

of 86.33%. Students who are predicted not to graduate

successfully have to work hard and be wary of

academic warnings or even dropping out of school.

Therefore the establishment of this model has

practical meaning.

4 CONCLUSION

As time goes by, more and more college students are

facing academic warnings or even dropping out.

However, at the same time, there are also many

college students who have achieved excellent results.

In order to explore the factors behind, this paper on the

one hand summarizes the previous studies, and on the

other hand adopts the method of correlation test to

explore the influence of some social or economic

factors on students' academic performance which have

not been involved in the previous researches, as a

supplement to the previous researches.

In this paper, the effects of five factors, namely,

age at enrollment, gender, GDP, course types, and

attendance time, on students' academic performance

were investigated. It was found that all five factors are

related to students' academic performance. The age at

enrollment is negatively related and GDP is positively

related. When students' attendance time is during the

day and their gender is female, they will have a more

significant advantage in their performance.

Multicategorical logistic regression analysis found

that gender has the most significant effect on students'

academic performance among these factors. By

examining these factors, students will be in a better

position to identify the causes that affect their

academic performance and then correct them, thereby

avoiding the consequences of dropping out of school.

The study in this paper also has its limitations. The

factors involved in this paper are still not

comprehensive enough, such as parents' education

level, the area where the students were before

enrolling in school and other factors have not been

explored. At the same time, the data samples are not

collected objectively and comprehensively, and fail to

include different regions and different peoples. At the

same time, the fitting effect of the logistic regression

model is not particularly good. The research done in

this paper is for posterity only, and the investigation

of the role of some factors may not be correct. Future

research can work on the objectivity and

comprehensiveness of data collection to make the data

more persuasive and of universal value. In addition to

this, there are many other possible factors that also

play a role in students' academic performance but have

not yet been mentioned in this paper and previous

studies. Future research can continue to explore the

effects of these remaining possible factors and find

models that are more suitable for fitting these factors,

so that the regression model can be more specific,

comprehensive, and able to more accurately predict

students' academic performance, which will have the

effect of helping them avoid dropping out of school.

REFERENCES

J. Li, M. Ou, “Intergenerational transmission of

poverty:Poverty level and academic achievement,”

Financial Science, 2018, 9, 107-119.

B. Q. Song, “Research on the relationship between college

students' classroom space preference and academic

performance,” Yangzhou:Yangzhou University Master

Thesis, 2020, 72.

Q. X. Wu, S. Cui, “Can't Student Leadership and Academic

Achievement Be Combined?-A study based on the

tracking survey data of college students in Beijing,”

Fudan Education Forum, 2019, 17(04), 71-79.

J. L. Wang, L. Yang, S. F. He, “A study on the relationship

between traditional values, classroom silent behaviour

and academic achievement,” Heilongjiang Higher

Education Research, 2019(9), 89-95.

G. Q. Zhang, M. Zuo, Q. Ma, “The influence mechanism of

microcourses on college students' academic

performance,” Statistics and Management, 2020,

35(02), 92-96.

C. Y. Li ,P. J. ,Lu, Y. D. Shang, “An investigation of the

impact of teacher credibility on college students'

academic performance,” Future and

Development,2020,44(10):97-104.

Z. J. Hao, C. Lin, H. Y. Cao, X. M. Chen, Y. F. Ding. “The

influence of social circle on college students' academic

performance,” Modern Education Technology, 2020,

30(10), 85-90.

Y. Q. Xiong, X. C. Niu, “Do college students' part-time jobs

affect their academic performance?-Evidence from

meta-analysis,” Fudan Education Forum, 2022, 20(01),

62-70.

M. Z. Song, T. T. Ai, H. Z. Fu, “The effect of college

students' psychological capital on academic

performance - Comparison and selection of competitive

models,” Higher Education Forum, 2021(03), 114-119.

J. Huang, Q. Chen, “The effect of class atmosphere on

college students' academic performance: the mediating

role of peer relations,” International Public Relations,

2021, 4, 44-45.

N. Allison, S. Michael, “Sleep Disparities and the

Academic Achievement Gap in 1.9 Million College

Students,” SLEEP, 2022, 45(1), A31.

Research on the Inﬂuencing Factors of College Students’ Academic Achievement Based on the Correlation Test Approach