Analysis and Prediction of Students’ Performance Using Machine Learning

Predicting students’ performance has drawn attainable interest in education. However, quantifying students’ performance is quite challenging as it depends on several factors. This study focuses on using educational data mining techniques (EDM) to estimate students’ final grades based on their social, school-related, and demographic data. Although past evaluations have a major impact on student achievement, however, explanatory analysis has identified that there are other important features (e.g., absences, study time, etc.) that can determine the students’ performance. So, some efficient feature selection techniques such as Random Forest Importance, ANOVA, and Recursive Feature Elimination, and dimensionality reduction methods such as Factor Analysis, PCA, and LDA are applied to the Portuguese and Mathematics lessons dataset. We have highlighted two modules comparison of both binary and four-level classification; firstly, we analyze the accuracy performance of different classical classifiers using all the dataset features, and secondly comparison with the relevant selected features. As a result of this research, more effective student prediction tools can be created, benefiting both the management of school resources and the quality of education.