Neutrosophy-Driven Deep Learning for Predicting Student Performance

Abstract

This paper proposes a hybrid architecture using several deep learning models in the 
neutrosophy environment for predicting student learning outcomes. The proposed framework 
proceeds on deep neural network models with the neutrosophy encoder/decoder. The 
experimental results on the actual data of the Hanoi Metropolitan University show that the 
proposed model's RMSE, MAE, and R2 metrics are better than the deterministic models. 
Especially, when using the previous 3 semesters for the prediction model, the R2-score of all 
05 proposed frameworks are approximately 60%, which proves the suitability of the proposed 
approach for the educational data and the prediction student outcomes problem. 

DOI: 10.5281/zenodo.15786779