A Comprehensive Score Prediction Method for College Students Based on Multi Model Fusion

Abstract

The rapid development of information and artificial intelligence technologies has brought new opportunities for innovation and transformation to higher education. The integration of these technologies with higher education teaching has led to innovative applications of data-driven decision support in teaching. This paper analyzes the characteristics of various machine learning models, comprehensively utilizes the strengths of each model, constructs a stacking model, and optimizes the model through Bayesian hyperparameter optimization to achieve accurate prediction of comprehensive grades for college students. The academic performance of students in a specific major in Central China is used as experimental data for model validation, achieving ideal predictive results. This method is characterized by a minimal feature set, high predictive accuracy, and strong stability. The comprehensive score prediction model, constructed through analysis of a selected specialty collection, holds significant guidance value in assisting teaching and research departments with curriculum and subject development.