Predicting Student Performance through Machine Learning Methods: Naive Bayesian Classifier

Education includes many aspects of knowledge, skills, impact on society, and values that shape critical thinking and personal development. Student outcomes are linked to a wide range of factors, which include teaching quality, parent involvement, socioeconomic status, and cultural dynamics. The economic conditions also affect education policy, funding, teacher salaries, and competition for human capital globally, as well as access to education in developing countries. Among these complexities, predictive models are important in identifying the at-risk students and optimizing the use of resources to foster better learning outcomes. This research work is focused on the prediction and classification of the performance of students in the Portuguese language in a cohort of 649 learners. In this study, the NBC model is used, together with two new hybrid models optimized by the SHO and IAOA. The performances of the models are critically evaluated to assess their effectiveness. The results reveal the better predictive performance of the NBSH hybrid model, outperforming the NBIA model in predicting G2 performance with a Recall of 0.897 against 0.886, which means an improvement of 1.23%. Besides, the NBSH model is still ahead in the prediction of G3 outcomes, reaching the highest Recall of 0.875. These findings indicate the possibility of hybrid models in educational data performance prediction, which will widen the avenue to informed intervention and resource planning in educational systems.