Multi-objective Hybrid Optimization-based Feature Selection for Sentiment Analysis

Reviews, ratings, and texts that customers have made online on items they have purchased online. These assessments assist business owners to assess their strengths and flaws and take action. The company owner cannot track each review due to the large volume of reviews. Recently, sentiment analysis for review summarization has demanded techniques from massive datasets. The sentiment analysis from the raw texts becomes a challenging task that needs to optimized feature selection approach. Using Natural Language Processing, the pre-processing step must eliminate the undesirable material from the text reviews (NLP). The extraction of more significant and reliable features for accurate sentiment analysis was the focus of the feature extraction phase. To that end, we present the Hybrid Multi-objective Optimization (HMO) method, which combines Particle Swarm Optimization (PSO) with the Krill Herd Algorithm (KHA). The HMO selects features from the supplied pre-processed reviews. The classification is performed by using different machine learning classifiers. The simulation results prove the efficiency of the proposed model compared to the existing solutions in terms of precision, recall, F1 score, and execution time.