Neuro-Fuzzy Decision System for Student Stream Selection

Abstract

Choosing an appropriate academic stream plays a crucial role for students in shaping their careers. However, choosing the right academic stream may be influenced by various factors related to uncertainty and subjectivity, such as academic performance, parent’s influence, socioeconomic background, career goals, and personal skills. A Neuro-Fuzzy Decision System (NFDS), which operates using Adaptive Neuro-Fuzzy Inference System (ANFIS), can assist in offering an intelligent recommendation for selecting a suitable student stream. Fuzzy logic coupled with learning capability of neural networks allows modeling uncertainties and incorporation of expert knowledge in addition to taking into account historical data on education. The study uses a dataset that includes features related to academic, demographic, socioeconomic, parental, and career information obtained through questionnaires. NFDS uses fuzzy membership functions, rules, and learning techniques to recommend one out of three streams – Science, Commerce, and Arts. Performance testing of ANFIS-based NFDS is evaluated against random forest, decision tree, xgboost, support vector machine, and artificial neural network algorithms. The suggested ANFIS method was able to reach the highest accuracy of 94.00%, along with having a precision of 94.74%, recall of 83.72%, F1 score of 88.89%, and AUC value of 0.957. Additionally, feature importance analysis highlighted that communication skills, decision-making freedom in the career path, financial aid, and social background were some of the key attributes when it comes to selecting the right streams. The findings indicated that the suggested neuro-fuzzy methodology is able to provide accurate recommendations.