Answers are grounded in this paper and cite their sources.
Modeling study demonstrates high-accuracy teaching evaluation in higher education courses, indicating the value of dynamic behavioral data and explainable machine learning.
Current evaluation of teaching effectiveness in higher education often relies on final grades or subjective student evaluations, which cannot adequately reflect dynamic learning behaviors or provide interpretable quantitative evidence. To address this problem, this paper proposes a data-driven quantitative evaluation model for higher education teaching effectiveness. First, four types of heterogeneous data from academic affairs systems, online learning platforms, classroom interaction tools, and assignment submission systems are fused to construct derived features such as learning engagement, knowledge mastery slope, assignment timeliness, and class participation activity. Second, a hierarchical prediction model is built using a weighted fusion strategy combining XGBoost and LightGBM. Third, the SHAP method is introduced to rank feature contributions and provide individual-level interpretability. The model is evaluated through five-fold cross-validation, feature ablation, and cross-course rotation testing. Experimental results show that the proposed ensemble model achieves 91.6% accuracy, with process-behavior features contributing substantially to prediction performance. The average cross-course accuracy reaches 86.7%, indicating good transferability. The study provides an interpretable technical framework for quantitative teaching-effectiveness evaluation.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: