Machine Learning Based Sentiment Analysis and Swarm Intelligence

Abstract

Social networking platforms, online news outlets, and weblog hosting services continue to expand, and with them come an increasing number of user-generated content contributions such product evaluations, comments on recent articles, and more. Products, movies, shopping sites, and review sites are common areas for customer feedback. The sheer volume and rate of growth of material that expresses opinions is becoming a burden on manufacturers who must manually categorise this data. Also, the perspective on entities at the level of aspects is expected by the public. It is for this reason that an automated sentiment analyzer must be built, one that can detect the bipolar and multipolar sentiment polarity of documents and/or aspects. People's ability to voice their opinions openly in public has greatly increased with the advent of various social networking apps. As a result, this helps to further the field of automated emotional analysis by providing a wealth of data on which to base analyses of people's feelings. User review categorization and analysis has emerged as an important part of sentiment analysis in recent years. Opinion mining is used to determine the degree of positivity or negativity in each user review posted on a social network. Numbers, star ratings, and descriptive text are the three polarity indications in a review. The sentiments of the public have been analysed using a wide variety of machine learning methods, but these methods often fall short in key areas such as classification accuracy, precision, recall, and F-measure due to pre-existing classification problems such as the two-class problem, overfitting, and parallel processing. The primary goal of the study is to create a fully automated system that can analyse a massive dataset of movie reviews using aspect-based SA or OM. We use natural language processing to tally up the good, bad, and ugly reviews. The research enhances advertising efforts and guides customers to the most suitable products. In this study, we use a variety of machine learning and swarm intelligence optimization techniques to the problem of determining the tone of movie reviews. Profits are increased and product failures are decreased thanks to this study for a wide range of businesses. The effectiveness of these procedures has been measured using MATLAB data from critical assessments of movies. The simulation results demonstrate that the proposed HIRVM scheme outperforms the state-of-the-art sentiment analysis schemes like HKELM, ID3, and J48 with respect to accuracy (96.82 percent), sensitivity (97.1 percent), specificity (91.2 percent), precision (96.2 percent), recall (90.2 percent), and F-Measure (89.5 percent). As compared to conventional methods, the suggested HIRVM significantly reduces both processing time (28.14s) and processing cost.