Research On Data Modeling Of Academic Performance Of College Students

Educational data mining in colleges and universities is a technical means of mining and analyzing large-scale and full-sample educational data in Colleges and universities.It has the value of governance in colleges and universities,which can help improve the decision-making ability,management efficiency and teaching effect of colleges and universities.Based on the technical means of educational data mining,to construct the data model of academic performance of college students and to excavate and utilize the value of college educational data is helpful to enrich the theory of student development and improve the system of student management and academic support in Colleges and universities.The research on academic performance of college students has a reliable theoretical basis for empirical research.This study absorbs the theoretical framework of academic performance of university students,including David Lavin's analytical framework of influencing factors of academic performance,Astin's theory of student engagement and I-E-O model.George Kuh's theory of student engagement and the model of success factors of University students,as well as the results of NSSE,CCSS,NCSS and other research programs.A four-dimensional analysis framework of influencing factors of academic performance is constructed,including school environment,social and demographic characteristics,personal characteristics of students,and student engagement.Based on this analysis framework,systematic data analysis and data mining are carried out to reveal the problems existing in the educational data system of case colleges and universities as well as the problems existing in the academic performance of case colleges and universities.This study adopts the method of case study,focusing on revealing the unique situation and problems of academic performance of individual college students.Through case analysis,it is also helpful to understand and analyze the common problems of academic performance of universal university students.Taking A University as an example,the study builds an integrated educational data system,collects the data of teaching management information system and the related data of students' academic performance for data cleaning and storage.Descriptive statistics and variance analysis are used to analyze the characteristics and differences of students' academic performance.With the data from teaching management information system,the objective data in the integrated educational data system and all the data in the integrated educational data system,from the three dimensions of social and demographic characteristics,personal characteristics and student engagement,the regression model is constructed by traditional statistical methods such as multiple linear regression and binary logistic regression.The classified models are constructed by Bayesian network,decision tree,artificial neural network,support vector machine and other educational data mining methods.Comparison on the validity of each model is done to figure out the model for predicting the academic performance of students.The results show that there are gender differences,regional differences and ethnic differences in academic performance in terms of social and demographic characteristics.The difference of parents' educational level will not bring about the difference of students' academic performance.From the dimension of individual characteristics,there are also disciplines differences in academic performance;there is a positive correlation between college entrance examination results and students'academic performance;personality,academic self-efficacy,psychological status,physical fitness test,psychological alert status,student identities as cadres and party members are all correlated with academic performance.From the dimension of student engagement,there is a significant positive correlation between students'academic performance and participation in academic lectures,voluntary services,associations and the use of library resources.There is a negative correlation between learning experience and students' academic performance.Academic goals are positively correlated with academic performance.By comparing 27 data models of regression and classification,it is found that the model based on all data in the integrated education system is the most effective.Multivariate linear regression model can explain 65.4%of the variance of students'academic performance;the explanatory power of social and demographic dimension variables is between 13%and 18%,the explanatory power of individual dimension variables is between 7%and 20%,and the explanatory power of students engagement dimension variables is between 10%and 17%.The highest prediction accuracy of binary logistic regression model is 69%.The 12 classification models based on Bayesian network,decision tree,artificial neural network and support vector machine are different in running time,prediction accuracy and sensitivity.Relatively speaking,the predictive effectiveness of Bayesian network classification model and support vector machine classification model is higher than that of decision tree classification model and artificial neural network classification model,and the stability of Bayesian network classification model is higher than that of support vector machine classification model.Ten independent variables were reduced by the reduction of independent variables,and the effectiveness of the model was not significantly reduced.