Grade Prediction of Student Academic Performance with Multiple Classification Models

Abstract

The achievement evaluation and development prediction of college students are the core of student management in universities. The traditional student evaluation method only focuses on the evaluation of students' past achievements, but lacks the prediction of students' future development. The change of grade prediction of students' future academic performance has great values for schools to strengthen education management. In this paper, Naive Bayes, Decision Tree, Multilayer Perceptron and Support Vector are utilized as classification models to predict students' academic performance. And an exhaustive comparative study is carried on the datasets of students' information provided by university of electronic science and technology. Among the models, multi-layer perceptron model has demonstrated powerful effectiveness, which achieved 65.90% accuracy on the training set and 62.04% accuracy in the test set. If the data sample is large enough, the experimental results will be more accurate.