.png){kind=icon}
.png){kind=icon}
Return
Complementary CatBoost based on residual error for student performance prediction
.png){kind=icon}
PRE
AI
DOI：10.1016/j.patcog.2024.111265.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Student performance prediction is crucial for early identification of potential students who may fail the final exam. In this paper, we propose a residual error-based Complementary CatBoost approach (C-CatBoost) for student performance prediction. Unlike the conventional CatBoost, our approach incorporates the evaluation of errors between the predicted values and target values. We design the residual error model to complement the performance of CatBoost. Besides, we integrate the CatBoost and the residual error model for the final prediction. The experimental results show that the C-CatBoost outperforms the comparing models in all the evaluated metrics. Specifically, our C-CatBoost achieves the lowest root mean square error of 1.1099 for predicting the final grades in Mathematics, surpassing the comparing models by 8.06% to 17.99% and achieves the value of 1.0246, outperforming the comparing models by 2.18% to 10.41% for the Portuguese language course. These promising results validate the effectiveness of complementary approach in enhancing student performance prediction, indicating the C-CatBoost could be a valuable tool within the educational institutions, contributing to the improvement of education quality.
Keywords:
Complementary CatBoost
Residual error
Model integration
Student performance prediction
Educational data mining
Journal
IF:
7.6
Papers:
1.3W
Citations:
4.5W
Organization
No organization information available