Machine Learning Based Admission Data Processing for Early Forecasting Students' Learning Outcomes

Abstract

In this paper, the authors explore the factors to improve the accuracy of predicting student learning outcomes. The method can remove redundant and irrelevant factors to get a “clean” data set without having to solve the NP-Hard problem. The method can improve the graduation outcome prediction accuracy through logistic regression machine learning method for “clean” data set. They empirically evaluate the training and university admission data of Hanoi Metropolitan University from 2016 to 2020. From data processing results and the support from the machine learning techniques application program, they analyze, evaluate, and forecast students' learning outcomes based on admission data, first-year, and second-year academic performance data. They then submit proposals of training and admission policies and methods of radically and quantitatively solving problems in university admissions.