Hybrid Fuzzy Rule-Based Classification System for MOODLE LMS System

Abstract

Educational data are widely applied to predict students’ academic performance in educational systems. Prior research mainly used students’ past learning data to predict their future performance. However, these predicted results could not provide teachers with the opportunity to remediate the students in time.
In order to achieve the effect of early warning, this study uses only the activity log of the first third of the semester to build models and prediction results. A hybrid classification decision mechanism is proposed to combine the results of different predictions based on the accumulated training cases to further improve the accuracy of prediction.
The proposed system is then applied to discover students’ learning outcomes in a C programming language course in the early stage of a semester according to the log files of the MOODLE LMS system.
The results show that the transformation of learning activity data has a critical impact on prediction accuracy. Using cumulative training cases can significantly improve prediction accuracy. And the proposed hybrid fuzzy classification decision-making scheme, which combines data conversion with cumulative training cases, can produce higher prediction accuracy by using just one-third of a semester’s learning activity data.