Comparative Analysis of Machine Learning vs. Soft Computing in Educational Outcome Prediction

Abstract

This study compares traditional machine learning (ML) methods—logistic regression, SVM, Random Forest—with soft computing (SC) techniques, including fuzzy rule-based systems, neuro-fuzzy networks, and genetic-optimized neural models, in predicting student academic outcomes. Using data from 1,200 secondary school students across Ghana, Australia, and Portugal, results indicate that SC models attain higher interpretability and comparable accuracy (±1.5%) relative to black-box ML, while offering clearer rule insights and greater robustness to noisy features.