International Journal of
Information and Education Technology

Abstract

In the current age, students' academic performance deterioration is a very crucial problem in engineering education. Prediction of low-performing students at an early stage is important so that their faculties and administration could provide timely support. The present study attempts to perform this prediction task at the entry-time with the help of four single supervised educational data mining algorithms, namely: Decision tree, Naïve Bayes, k-Nearest Neighbor, and Support Vector Machine along with an ensemble method called “Random Forest”. These classifiers have been applied to a students‟ dataset of an Indian Engineering College, having four categories of parameters viz., student‟s background, academic, social, and psychological parameters. Different libraries of Python programming language such as Pandas, Seaborn, Scikit-learn, and Scipy were used for analysis, visualization, classification, and statistics computation, respectively. The present study shows that among all of the five algorithms, Naïve Bayes gives the highest accuracy with 89%, and finally to improve the results, a model is proposed in which three Naïve Bayes classifiers were integrated with the help of 'Bagging'. The achieved accuracy with the proposed model was 91%, with the highest recall and highest precision for identifying low performers.