Data Mining Research Of College Students’ Academic Performance And Graduation

With the popularization of higher education and the strong financial investment of government,the educational level of Chinese institutions of higher learning has been improved.At the same time,the rapid development of technology has provided convenience for school administrators,teachers and students.In the information age,college students have left a lot of information in the process of receiving education.By mining the knowledge and patterns hidden behind the data information,it provides a strong support for improving the management efficiency of managers and can better serve for students.Taking the data of students’ academic performance and graduation destination which are most closely related to students as the research object,this paper use technologies of data mining to mine both of them,which provides a new space for the application of data mining technology in the field of education.In view of the application needs of technologies of data mining in college students’ academic performance and graduation destination,this paper takes 5445 graduates from 68 majors of a university in 2018 as the research object,and uses the clustering and classification model in data mining as the technical means to conduct in-depth data information mining research.The main research work and results are as follows:(1)Build a variance analysis model of students’ achievements based on graduation destination.First of all,the original achievements data and graduation data were preprocessed.According to the attributes of their majors,students in 68 majors of the school were divided into five major categories.They are namely,science and technology,agriculture,economics,management and liberal arts.Secondly,through descriptive analysis,the total academic scores of students in five majors were ranked as follows: liberal arts > economics> ariculture> management> science and engineering.The graduation destinations of students are in the order of private enterprises,admission to higher education,state-owned enterprises,flexible employment,waiting for employment,party and government organs,public institutions and foreign-funded enterprises.Finally,the graduation destinations were grouped with the following categories: further study,employment outside the system,employment within the system and uncertain employment,and the total scores of 68 majors and the scores of the first 7 semesters were analyzed by means of variance analysis.The results showed that there were significant differences in the total scores of students in 59 majors,among which there were significant differences in the scores of the first 6semesters.Academic achievement is still an important factor affecting the choice of graduation destination,which provides theoretical support for subsequent data mining.(2)Build a cluster analysis model based on student achievement.K-Means and FCM algorithm in data mining are adopted to construct the clustering modesl to cluster 4939 students in59 mijors by taking students’ grades of the first six semesters as eigenvalues and selecting 5 as mumbers of cluster.The two models are compared with the evaluation index of clustering effect,and it is concluded that the clustering model based on FCM is more suit able for the clustering analysis of students’ grades.It has the highest overlap with the data of 5 categories of students according to their major categories.The students in each category output based on FCM clustering model are analyzed to explore the characteristics of students in each cluster category.(3)Constructing the classification prediction model of college students ’ graduation destination.Aiming at solving the problmes of lackness of forecasting model,lower accuracy and required characteristic variables of model,this paper take five classification categories after clustering,the first semester grade six students and weighted average grade as eight variables characterized tag classification to establish support vector machine prediction model.This model selects gaussian function as the kernel function,and the 1-V-1 method was selected to carry out the multi-classification experiment of SVM.The indexes were compared with the classification and prediction models established by ANN and RF methods.It was found that the graduation destination prediction model based on SVM had the highest accuracy,and it was superior to the classification and prediction model based on the original five major categories.At the same time,the importance of each factor on the graduation destination is calculated.The results show that the fifth and fourth semester scores have a greater impact on the graduation destination,which is worthy of education administrators and teachers to invest more energy in this stage.