With the continuous development of digital campus construction in colleges and universities, the campus big data environment gradually tends to be complete. The data of students’ behavior characteristics can directly reflect students’ daily behavior habits and is related to students’ performance. Therefore, it is necessary to mine the data of students’ behavior characteristics accurately and systematically to establish a high-precision achievement prediction model and provide reliable decision-making basis for student management. This paper presents a model of optimized extreme learning machine based on particle swarm optimization (PSO-ELM). The data derive from Yi ban platform data and daily statistical data. Firstly, filter the data of students’ behavior and process its vacancy value then analyze the correlation of the factors affecting students’ performance by binary correlation analysis. After that, establishing an optimized extreme learning machine (ELM) model predict students’ performance. To verify the model, data were randomly assigned to the training group and the test group that used for validation according to the ratio of 3:2. The results indicate that the model has good performance and precise prediction accuracy.