Mamdani Fuzzy Model for Learning Activities Evaluation

Abstract

The intent of this paper is to determine the extent to which fuzzy model could suitably modelled learner activities in Elearning system. However, the paucity of public dataset that meet the exact requirement of this work poses challenges, which necessitate dataset simulation. The detail approach used for the dataset simulation and the fuzzy model were discussed. Construction of the Inference Mechanism using the Relational Calculus and Mamdani approaches were demonstrated. The performance of the simulated model in MATLAB was measured using classifier uncertainty and confusion based metrics. The Mean Absolute Error (MAE) is 10.45; Root Mean Square Error (RMSE) is 8.71. The result shows that Fuzzy logic (White-Box Model) has a low classification error and invariably a higher accuracy for estimating learner activities. Subsequently, the result obtained shall be revalidated using live data of students' activities in an online course. Furthermore, the current Mamdani's model performance shall be compared with its equivalent Neuro_Fuzzy Model. The more efficient of the two models shall be the choice for integration into an Open Source Learning Management System for automatic learning activities evaluation. General TermsStudent Modelling, Soft Computing.