The multi-parameter optimized belief rule base for predicting student performance with interpretability

TL;DR

This paper introduces an interpretable student performance prediction model using a multi-parameter optimized belief rule base to improve accuracy and clarity.

Contribution

The novel contribution is an interpretable student performance prediction model with multi-parameter optimization and attribute selection.

Findings

An attribute selection method based on random forest improves feature relevance for student performance.

The model incorporates interpretability criteria during optimization to maintain clarity.

Case study results verify the effectiveness of the proposed method in real-world educational settings.

Abstract

Predicting student performance is essential for making informed teaching decisions, customizing learning, and ensuring educational equity. When developing student performance prediction models, it is crucial to provide high prediction accuracy, a clear and logical prediction process, as well as easily understandable and traceable prediction outputs. The Belief Rule Base (BRB) combines expert knowledge to ensure accuracy while also having a certain degree of interpretability. However, the following problems still exist: When there are too many attributes, BRB will encounter the problem of rule combination explosion; After the optimization stage of the BRB model is completed, its interpretability may decline. Furthermore, when experts have limited knowledge, the reference values they cite may weaken the prediction accuracy of the model. In response to the above problems, this paper presents an interpretable student performance prediction model based on a multi-parameter optimized belief rule base(IBRB-m). Firstly, an attribute selection method based on random forest was introduced to screen out the important features that affect students’ academic performance; Secondly, The criteria for interpretability in the model optimization process have been defined. Finally, a student performance prediction model is constructed and a model parameter optimization method with multi-parameter optimization and interpretable constraints is proposed. The effectiveness of this method was verified through a case study of the performance of students in a certain school.

Figures