Abstract:

Based on the analysis of different metrics, this research identifies the most performing predictive algorithms in educational data environment using the Faculty Support System (FSS) model, and provides a summary of current practice and guidance on how to evaluate educational models. FSS model is implemented in different phases with usage of different algorithms for an early prediction of students’ failure and help them to enhance their performance supported by Centre of Excellence team and complete the course with a good grade. This study uses Naive Bayes, Multiple Layer Perceptron, and Random Forests and J48 decision tree induction to build predictive data mining models on 111 instances of students’ data. We applied 10-fold cross-validation, percentage split and training set methods on data and performance metrics were used to evaluate the baseline predictive performance of the classifiers. The comparative analysis shows that the Multiple Layer Perceptron performed best with accuracy of 82% and Random Forests came out second with accuracy of 79%, J48 and Naïve Bayes came out the worst with accuracy of around 60%. The evaluation of these classifiers on educational datasets, gave an insight into how different data mining algorithms predict student performance and enhance student retention.

Keywords:

Educational Data Mining (EDM), classification, naïve Bayes, decision trees, random forests, support vector machines, neural networks.