Students Academic Performance Prediction Based on Square Root Data Transformation and Ensemble Technique

Abstract

Data mining research is evolving rapidly in the educational sector because of the vast amount of student information used to detect and explore useful patterns applicable to student learning behaviour. Predicting students' progress is an essential task in any educational institution. To assess student performance, educational institutions may use educational data mining to improve their teaching practices and learning processes. All these modifications lead to enhancing the success of students and overall academic results. In data mining, classification is a popular technique that has been widely tested out to find student outcomes. An approach based on data transformation and the Ensemble method to predict student success is suggested in this report. The efficacy of the student's predictive model is measured using several classifiers: ErrorCorrecting Output Code (ECOC), K-Nearest Neighbour (KNN), Ensemble, Naïve Bayesian (NB), and Decision Tree (DT). The results obtained by training the different classifiers with square root transformed features improved the classification accuracy from 83% to 86%, thus improving the performance prediction model's overall performance. For the X-API dataset, this suggested technique also created a better prediction accuracy than related works that used the same dataset.