A Systematic Approach to Transform Machine Learning Students’ Performance Prediction Model into Preventive Procedures

Abstract

The modern day educational institutes are craving novel procedures to utilize the students’ data to amplify their prestige and improve the education quality. One of the major problems an instructor/institute experiences is the thorough monitoring of students’ academic progress, in a course, and instigate preventive procedures to offer additional support to the students with unsatisfactory academic progress. Educational Data Mining tools, specifically Machine learning classifiers, appear supportive to develop prediction models which forecast students’ final outcome in a course. This research evaluates the effectiveness of machine learning classifiers to monitor students’ academic progress and informs the instructor about the students at the risk of producing unsatisfactory final result in a course. The dataset is pre-processed with Pearson correlation feature selection algorithm to discover the features which influence the students’ academic performance. A set of machine learning models are developed and compared through accuracy, sensitivity, specificity, F-measure and Mathew Correlation Coefficient, to choose the finest model. J48 decision tree prevails other models by achieving accuracy and F-measure of nearly 0.90. Simple logistic appeared the least effective model while an altered version of k-nearest neighbor achieved highest sensitivity but remain ineffective due to lower accuracy and F-Measure. The ideal model is further transformed into easily explicable format and then interpreted into a set of supportive measures to carefully monitor students’ performance from the very start of the course and a set of preventive measures to offer additional attention to the struggling students.