Subsethood-based Fuzzy Rule Models and their Application to Student Performance

The focus of this paper is the use of fuzzy approaches to classify student academic performance, which so far has not been performed satisfactorily by existing fuzzy techniques. Instead of using methods that solely rely on expert opinions, student performance evaluation is herein conducted using fuzzy rule-based models which combine expert knowledge with knowledge extracted from data. Significant advantages of the present work are shown by comparing the results obtained with various alternative techniques. In addition to the ability to produce classification that helps the students to understand their performance, the use of membership value degrees in both rule antecedents and conclusions allows one to confirm or refute results in certain borderline cases that were obtained by other means (e.g. by a human evaluator). In particular, this advantage is coupled with a simpler modelling mechanism that minimizes human intervention by avoiding the use of preset threshold values which are typically used in conventional subsethood based models.