A COMPARATIVE STUDY OF BASE CLASSIFIERS IN PREDICTING STUDENTS’ PERFORMANCE BASED ON INTERACTION WITH LMS PLATFORM

Abstract

E-learning platforms known as Learning Management System (LMS) generate huge amount of data that need to be examine in order to derive meaning out of it. This can be achieved using data mining techniques on large educational data, this is a field also known as Educational Data Mining (EDM). One popular application of EDM is prediction of students’ performance. This application seems to be difficult due to the diverse nature of the variables that affect performance of students such as culture, family background, psychological history, previous academic performance, parents’ economic situation, and previous schooling. In this paper, data mining technique was used to predict students’ performance based on their interaction on a LMS. The LMS used is the Modular Object Oriented Dynamic Learning Environment (MOODLE). Datasets containing students’ activities on MOODLE were used for the study. The base classifiers used for comparison are: Decision Tree (DT), Naïve Bayes (NB) and K-Nearest Neighbor (KNN). Waikato Environment for Knowledge Analysis (WEKA) was used for data preprocessing, attributes selection evaluation, result analysis and 10-fold cross validation. The results obtained indicates that DT is the best model with 84.1% accuracy which outperforms NB and KNN with accuracies of 83.7% and 76.7% respectively. A correlation analysis showed that the assignment submission attribute was identified as the most significant feature that have the most impact on the prediction of students’ performance.