The Research of Predicting Student's Academic Performance Based on Educational Data

In recent years, with the continuous improvement of teaching informatization, online teaching or online and offline hybrid teaching has become the new normal of teaching in some schools. However, the biggest problem in online teaching is that it is difficult to predict students' academic performance. Therefore, it is necessary to design an effective method to predict students' academic performance more accurately. In this paper, a student academic level prediction method based on stacking model fusion is proposed. Logistic regression, random forest, XGBoost, and Naive Bayes are selected as base learners according to optimal fusion criteria and model properties. Furthermore, the structure and distribution of the features of the dataset are optimized by data preprocessing, feature coding, and feature selection, and the upper limit of the model expression is effectively raised. On this basis, according to the features of dataset and model performance, we select the appropriate model for model fusion, and further improve the prediction effect. Experiments are conducted on OULAD and xAPI datasets, and the results show that the prediction accuracy of the proposed method is better than that of traditional prediction methods. Finally, we analyze the factors that affect academic performance and give some specific suggestions.