SA-FEM: Combined Feature Selection and Feature Fusion for Students' Performance Prediction.

Area of Science:

• Educational Technology
• Computer Science

Background:

• The COVID-19 pandemic necessitated a shift to online learning, highlighting challenges in maintaining educational quality.
• Existing online learning performance predictors often overlook the crucial internal correlations between diverse e-learning behaviors.
• The e-learning behavior classification (EBC) model offers a way to capture these inter-behavioral relationships.

Purpose of the Study:

• To propose a new online learning performance prediction model, SA-FEM, that accounts for the internal correlations of e-learning behaviors.
• To enhance the accuracy and reliability of online learning performance prediction through adaptive feature fusion and selection.
• To provide real-time monitoring and feedback for online learners by improving prediction models.

Main Methods:

• Development of the SA-FEM model utilizing adaptive feature fusion and feature selection techniques.
• Employing a fine-grained differential evolution algorithm to explore the feature space.
• Integrating the EBC model to guide adaptive feature fusion based on performance categories.

Main Results:

• The fine-grained differential evolution algorithm effectively identifies a robust feature space for online learning prediction.
• Adaptive feature fusion, guided by the EBC model, significantly improves prediction accuracy.
• Experimental results demonstrate that SA-FEM outperforms established benchmark methods in online learning performance prediction.

Conclusions:

• The SA-FEM model, incorporating adaptive feature fusion based on the EBC model, offers a superior approach to online learning performance prediction.
• Accounting for the internal correlations of e-learning behaviors is key to developing more effective predictive models.
• This research contributes to improving the quality and effectiveness of online education through advanced data analysis.