An Explainable PSO Based Feature Optimization Framework for Sentiment Analysis of MOOC Learner Feedback

Abstract

Massive Open Online Courses (MOOCs) of learner feedback in sentiment analysis is improving the essential quality and effectiveness using feature based optimization. Traditional approaches based on TF–IDF and standard machine learning classifiers are with challenges such as high dimensional feature spaces, limited interpretability, and suboptimal feature representation are restricting the analytics in the education platforms. This Proposed paper suggests XHOS-SA (Explainable Hybrid Optimization-based Sentiment Analysis), a novel framework that integrates Particle Swarm Optimization with an explainable feature weight optimization mechanism for sentiment classification. The traditional feature selection methods that mainly focus on feature reduction by eliminating irrelevant features, the proposed method XHOS-SA optimizes feature weights while preserving semantic relevance, thereby it enhances both classification performance and model transparency. The XHOS-SA incorporates explainable AI mechanisms by providing feature importance analysis, enabling to capture meaningful interpretation for educators and institution platforms helps to understand in the contribution of sentiment predictions. The Experimental process on a real-world MOOC learner review dataset shows that the proposed approach XHOS-SA significantly achieves an improvement than the TF–IDF model and traditional classifiers in terms of classification accuracy, precision, recall, and F1-score by the optimized feature weighting which leads to faster convergence and improved generalization performance. The results confirm that the proposed method XHOS-SA effectively balances predictive performance and improvement in the explainability, making it a robust and reliable solution for educational data mining, learner feedback analysis, and intelligent decision support systems in online learning environments.