Construction of Student Behaviour Prediction Model in Colleges and Universities under the Background of Big Data

Analyzing and predicting the behavior of students in colleges and universities is very important for the growth and development of students and for teachers to test their teaching results. In this paper, after using DT-kmeans clustering algorithm to analyze the density of students’ behavioral features, the center point of the class clusters is selected and it is iteratively updated to realize the cluster analysis of students’ behavior. Then the entropy weight method is used to calculate the weight of each behavioral feature to realize the hierarchical processing of behavioral features. Finally, a student behavior prediction model is established based on the K-nearest neighbor nonparametric regression method, and the accuracy of the model behavior prediction is optimized by the TPE hyperparametric method. The AUC value of the model constructed in this paper for student behavior prediction reaches 0.813 on average, and the model validity is verified. There is a positive correlation between the time students stay in the building and their academic performance, and the predicted grades of students who spend more time in the building are in the range of 80-100. And there is no significant difference in academic behavior among students with different spending behaviors. This paper lays a foundation for accurate prediction of student behavior in colleges and universities, and provides a reference for college and university teaching administrators to make scientific management and decision-making.