The Analysis and Prediction Mode of Students' Academic Performance and Social Behavior in the Background of Big Data

The popularization of digital campus brings convenience to the life of teachers and students, and provides rich data for educational data mining (EDM). EDM uses data mining and machine learning techniques to analyze student data, aiming to provide quality services. This paper uses big data to analyze students' behavior, constructs the academic performance prediction model, and provides a reference for decision-making for teaching assistants. The Dis_GAT network based on graph neural network is proposed and compared with traditional methods to prove its effectiveness. By preprocessing and analyzing student data, the behavioral characteristics were extracted, and identifying abnormal students using K-means and isolated forest algorithms to help the counselors fully understand the student situation. Finally, the Dis_GAT network prediction framework is proposed, transforming the academic performance prediction problem into the node classification problem in the graph structure, introduces the concept of graph neural network, improves the graph attention network, and considers the influence of early performance similarity on later academic performance.