An essential component of individual and society growth is student education. It could involve innovative curriculum design, efficient teaching techniques, educational technology, and resources. The majority of educational data mining (EDM) research, however, has concentrated on identifying at-risk children so that early, focused interventions can be given, as well as forecasting students' future performance. EDM seeks to create techniques for examining the distinct and progressively larger amounts of data produced by educational environments in order to gain a deeper comprehension of both learners and the environments in which they are taught. It is applicable to study the consequences of educational support and forecast how students will learn in the future. Using student performance datasets, we assessed the suggested feature selection technique's efficacy employing using five machine learning classifiers: Decision Tree (DT), Naïve Bayes (NB), Support Vector Machine (SVM), K-Nearest Neighbour (KNN), and Logistic Regression (LR). This suggested approach uses the chosen algorithm to verify training duration, analysis, recall, accuracy, and precision. But in order to determine which algorithm possessed the greatest best accuracy, they compared all of them the outcomes of this review will help academic researchers, practitioners, and professionals deal with imbalanced classification, particularly within the field of higher education