Using Educational Data Mining to Predict Student Academic Performance

An educational institution's primary objective is to create a learning environment that enhances student academic success by mitigating academic failure and promoting higher performance. In order to accomplish this, the institute needs an effective mechanism for quickly identifying students’ performance, in particular students at the risk of falling behind or failing a course. Using the classification approach of educational data mining, this study utilizes student descriptive, behavioral, and attitudinal data to predict academic performance at an early stage during a semester. Specifically, this study makes use of ruled-based, decision tree, function-based, lazy, multilayer perceptron, and probabilistic classification techniques for early student performance prediction. The models generated by several classifiers exhibited good performance with the model generated by the Random Forest classifier exhibiting an accuracy of 93.40% and a Kappa score of 0.9160. The experimental results of the study indicate the effectiveness of using a set of descriptive, behavioral, and attitudinal attributes to predict student performance at an earlier stage during the conduct of a semester