Predicting Student Performance Using Subtractive Clustering Fuzzy C-Means and Multiple Linear Regression

A common method for measuring student performance is to average all assessment component results. However, the average score does not accurately reflect students’ true performance, since each component assesses different learning outcomes related to knowledge, skills, and attitudes. To overcome this limitation, this study suggests a systematic approach for evaluating students’ academic achievement and behavioral attributes using a hybrid model that combines soft computing and statistical techniques. Several methods were developed and compared, including Fuzzy Inference System (FIS), Adaptive Neuro-Fuzzy Inference System (ANFIS), Fuzzy C-Means (FCM), Subtractive Clustering (SC), entropy-weighted FCM, and Linear Regression (LR). A hybrid model combining Subtractive Clustering Fuzzy C-means with LR using two independent variables (ScFCMLR2) was proposed. Experimental results on 415 student records showed that ScFCMLR2 achieved the highest accuracy (85%) and F-score (82.7%) among the models. The results demonstrate the reliability of the proposed model in evaluating overall student performance across both academic and behavioral attributes.