Improving Sentiment Analysis of Text Messages using Swarm-based Feature Selection and Deep Learning Techniques

Abstract

Natural language processing (NLP) strategies are increasing in importance, due to the massive growth of unstructured textual data in digital communication, especially on platforms like WhatsApp. Sentiment analysis is one such crucial factor that helps to extract valuable insights from this information overload. This research applies a sophisticated approach for sentiment classification of WhatsApp group chat data, by employing an ensemble swarm-based feature selection methods and deep learning architectures. This recommended approach synergistically integrates filter based feature selection metrics and deep learning hyperparameter tuning to improve both the accuracy and efficacy of sentiment categorization. Thorough testing with several datasets demonstrates the usefulness of this system compared to conventional approaches, with remarkable improvements noticed thereof. The outcomes provided evidence for its effectiveness and showed that it can outperform some well-established methods used in complex area like sentiment analysis in WhatsApp group chats.