Optimizing Multinomial Naïve Bayes in Sentiment Analysis through Hybrid Ensemble Techniques

Customers are vital in any business, understanding their needs will lead to a successful business presence, especially in today's age where all big businesses are connected online and use e-commerce platforms. However not all businesses use the data available and leverage for decision making. The study will focus on e-commerce data, specifically the customers' comments on products using Multinomial Naive Bayes to better understand customers. The Multinomial Naive Bayes (MNB) algorithm has been widely used for classification and sentiment analysis due to its efficiency and simplicity. However, the assumption of feature independence limits its performances, which fails to capture the complex contextual relationships inherent in natural language. This research proposes a novel idea and approach to enhance MNB algorithm for sentiment analysis by integrating it with hybrid ensemble techniques. The proposed methods of study combine MNB with advanced machine learning models, including logistic regression and transformers, to form a more robust hybrid ensemble. The ensemble utilizes a meta-learning framework to combine with other predictions of each model. Additionally, it incorporates techniques like feature engineering, n-grams and external sentiment lexicons