Analysis of Factors Affecting Academic Performance - Based on Machine Learning Methods

Abstract

Students' academic performance is the core indicator of educational quality. To overcome the limitations of traditional methods in analyzing the interaction and dynamic effects of complex factors, this study employs machine learning (logistic regression, ensemble learning, neural networks, SHAP value analysis, etc.) to integrate multi-source educational data and systematically explore the influencing factors and mechanisms of academic performance. Key findings: (1) Individual initiative (motivation, self-efficacy, strategy) contributed the most (48%), which was higher than that of teaching management (32%) and environmental factors (20%); (2) Motivation and self-efficacy have a synergistic enhancing effect, while working more than 30 hours per week weakens the role of the knowledge base. (3) There are significant differences among disciplines (mathematics emphasizes cognitive strategies, while medicine is regulated by psychological states). The research breaks through the static attribution paradigm, providing data support for precise teaching and pointing out the future direction of deepening dynamic modeling and interdisciplinary integration.