User Behavior Analysis and Prediction Model Construction in Higher Education Management Information Systems

Abstract

The education management system is an important tool for universities to manage academic information and information of staff and students. The article constructs a user behavior analysis model based on machine learning and a user behavior prediction model based on LR-XGBoost to analyze and predict user behavior in the educational management information system of colleges and universities. Specifically, after the collection of user behavior data in the university education management information system, descriptive statistics and cluster analysis are performed. Then, the relationship between user behavior and dropout/leaving is explored by the LR-XGBoost model, and the prediction performance is tested by comparing the prediction accuracy of the prediction model with other models. Cluster 3 users had significantly higher means on 12 behaviors in the college education management information system than those in clusters 2 and 1. Among the 12 behaviors, only closing web pages had a p-value greater than 0.05 and did not pass the z-test. The LR-XGBoost prediction model had the best performance with an accuracy of 95.51% and 94.53% for behavioral prediction and anomalous behavior prediction, respectively, and F1 values of 92.34% and 95.33%, respectively. It has a clear overall advantage in predicting user behavior for the next 1-10 days.

Keywords

• machine learning
• lr-xgboost model
• cluster analysis
• behavior prediction
• education management information system
• 97m80