Using educational data mining techniques to increase the prediction accuracy of student academic performance

Abstract

PurposeThis paper aims to evaluate educational data mining methods to increase the predictive accuracy of student academic performance for a university course setting. Student engagement data collected in real time and over self-paced activities assisted this investigation.Design/methodology/approachClassification data mining techniques have been adapted to predict students’ academic performance. Four algorithms, Naïve Bayes, Logistic Regression, k-Nearest Neighbour and Random Forest, were used to generate predictive models. Process mining features have also been integrated to determine their effectiveness in improving the accuracy of predictions.FindingsThe results show that when general features derived from student activities are combined with process mining features, there is some improvement in the accuracy of the predictions. Of the four algorithms, the study finds Random Forest to be more accurate than the other three algorithms in a statistically significant way. The validation of the best-known classifier model is then tested by predicting students’ final-year academic performance for the subsequent year.Research limitations/implicationsThe present study was limited to datasets gathered over one semester and for one course. The outcomes would be more promising if the dataset comprised more courses. Moreover, the addition of demographic information could have provided further representations of students’ performance. Future work will address some of these limitations.Originality/valueThe model developed from this research can provide value to institutions in making process- and data-driven predictions on students’ academic performances.