A HYBRID METAHEURISTIC FEATURE SELECTION FOR TEXT IN MOOC DATASET USING GENETIC ALGORITHMS AND PARTICLE SWARM OPTIMIZATION

Abstract

High-dimensional textual data in MOOC platforms, such as student reviews, discussion and comments, generates redundant and noisy features may cause significant challenges for the machine learning models that degrades model performance. To perform the essential and Effective feature selection to enhance the accuracy and interpretability in the machine learning models. This paper explores the application of two popular metaheuristic algorithms—Genetic Algorithms (GA) and Particle Swarm Optimization (PSO) for effective feature selection which selects the informative features from MOOC review textual data represented through TF-IDF. In this research work, GA is first employed to perform global exploration and identify diverse candidate subsets of features, while PSO is applied to refine and converge towards optimal feature subsets. In Addition, the results of both algorithms are integrated using a weighted scoring mechanism to construct a final hybrid feature selection that combines the importance scores from GA and PSO outcomes to leverage algorithms strength. Experiments on a MOOC review dataset is illustrated the hybrid strategy yields to superior performance compared with individual GA and PSO approaches using multiple classifiers, including Logistic Regression, Support Vector Machine, Random Forest, Multinomial Naïve Bayes, and Passive Aggressive Classifier. This paper contributes a development of a novel GA–PSO hybrid framework for text feature selection in educational data, a comprehensive experimental validation across multiple machine learning models and evaluation metrics, and demonstration of the practical utility of the selected features in enhancing automated learner feedback and performance prediction in MOOC environments.  This paper suggest that hybrid metaheuristics approach provides a promising possibility for effective feature selection in educational natural language processing (NLP) tasks and also facilitating improved analysis of learner-generated text.