Evolutionary Multiobjective Feature Selection for Sentiment Analysis

Sentiment analysis is one of the prominent research areas in data mining and knowledge discovery, which has proven to be an effective technique for monitoring public opinion. The big era with a high volume generated by variety sources provided enhanced opportunities utilizing sentiment various domains. In order take best advantage accurate analysis, it essential clean before as irrelevant or redundant will hinder extracting valuable information. this paper, we propose hybrid feature selection algorithm improve performance tasks. Our proposed approach builds binary classification model based on two techniques: entropy-based metric evolutionary algorithm. We have performed comprehensive experiments different domains using benchmark dataset, Stanford Treebank, real-world dataset created World Health Organization (WHO) speeches regarding COVID-19. shown achieve significant improvements both datasets, increasing accuracy all utilized machine learning text representation combinations. Moreover, achieves over 70% reduction size, provides efficiency computation time space.