With the rapid growth of educational data, how to extract valuable information from multi-source heterogeneous data has become an important challenge in educational research. The existing technology applications are scattered and lack systematic integration, which makes it difficult for educational practitioners to comprehensively evaluate the applicability of different methods. This paper systematically reviews the research progress in student behavior analysis, academic prediction and educational intervention from four perspectives: cluster analysis, prediction model construction, association rule mining and data visualization. By integrating domestic and foreign empirical cases, it is found that: clustering algorithms can effectively divide student groups, but the initial center selection and mixed data type processing still need to be optimized; deep learning models perform well in dynamic prediction, but the lack of interpretability limits the direct application of educational decision-making; association rule mining reveals the complex interaction between cognitive style and learning behavior; data distillation bridges the gap between algorithm output and educational interpretability through visualization technology. This paper proposes an innovative path for technical collaboration and emphasizes the importance of interdisciplinary cooperation in solving ethical challenges such as data integrity and algorithm fairness.
Qihao Chen (2025) studied this question.