PREDICTION ACCURACY ANALYSIS OF MACHINE LEARNING CLASSIFIERS ON STUDENT COURSE ASSESSMENT METHODS

Abstract

There is growing need to improve the quality of education through an effective service delivery from educators. Also, educational institutions are searching for ways to reduce student failure rate. The rapid growth in size and availability of student data and robust algorithms to generate machine learning models, more accurate predictions and tailored learning interventions can be factored. The research investigates the prediction accuracy of machine learning algorithms, including Logistic Regression, Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naive Bayes, XGBoost, and Gradient Boosting, applied to student learning attributes and course assessment. The aim is to evaluate the effectiveness of these algorithms in predicting student performance on various metrics. A dataset encompassing student learning attributes and assessment modes were analyzed. Each algorithm's predictive capabilities was assessed using accuracy, precision, recall and F1-score metrics. Logistic regression had the highest accuracy score of 0.93, SVM and XGBoost both achieved an accuracy 0f 0.90 while Random Forest, KNN and Naive Bayes achieved same accuracy score of 0.88 while Gradient Boosting achieved an accuracy score of 0.85 each which was the lowest. RF, SVM, KNN got the same F-score, recall and precision of 0.93, 0.97 and 0.90 respectively while Naive Bayes, XGBoost, and Gradient Boosting achieved the same recall of 0.94 while KNN had a recall of 0.97. Gradient Boosting had a precision of 0.89, and an F-score of 0.92, the F-score of Naïve Bayes was 0.93. This research underscores the potential of advanced machine learning techniques in enhancing educational outcomes.