A student academic performance prediction model based on the interval belief rule base.

Area of Science:

• Educational Data Mining
• Artificial Intelligence in Education
• Machine Learning for Student Success

Background:

• Student performance prediction (SPP) is crucial for timely educational interventions.
• Traditional belief rule base (BRB) models face challenges like attribute explosion and limited expert knowledge integration.
• Developing robust SPP models is essential for improving educational outcomes.

Purpose of the Study:

• To propose a novel SPP model, the interval belief rule base with random forest attribute selection (IBRB-C).
• To enhance the accuracy and efficiency of student performance prediction.
• To overcome limitations of traditional BRB models in handling complex datasets and expert knowledge.

Main Methods:

• Utilizing Random Forest (RF) for attribute selection to mitigate attribute explosion.
• Employing an interval BRB structure to manage uncertainty in predictions.
• Integrating expert knowledge with the Kmeans++ algorithm for parameter determination.
• Optimizing the IBRB-C model using the P-CMA-ES algorithm.

Main Results:

• The proposed IBRB-C model achieved a Mean Squared Error (MSE) of 0.0024 for graduate applications and 0.1014 for GPA.
• Ablation experiments validated the effectiveness and rationality of the IBRB-C approach.
• Comparative experiments confirmed the superiority of the IBRB-C model over existing methods.

Conclusions:

• The IBRB-C model offers a significant advancement in student performance prediction.
• This approach effectively balances expert knowledge with data-driven attribute selection.
• The model's high accuracy demonstrates its potential for practical application in educational settings.