Student Achievement Analysis and Prediction Based on the Whole Learning Process

Abstract

Blended learning is increasingly used in college teaching, and formative evaluation has become the main method for assessing student performance. Based on the formative evaluation data of an existing course, how to model, analyze and predict the possible problems of students in the future learning process and give recommendations on learning strategy are problems worthy of in-depth study. In this paper, Apriori algorithm was used to perform association analysis on the formative evaluation data of the Fundamentals of Programming course in Nankai University, the results indicate that there are strong association rules between SPOC video scores, case study assignments scores, etc. K-Means algorithm was used to perform cluster analysis on SPOC platform scores, offline course scores and final exam scores, the results indicate that the advantages and disadvantages of students of different categories are consistent in two semesters. Finally, the clustering results of the first semester were added to the data set, Random Forest was used for feature selection, and four ensemble learning models were trained respectively to predict final exam grades. The results show that the XGBoost model works best, the accuracy of predicting the final exam grades of two semesters is 77.02% and 80.10%, respectively.