Comparative Analysis of Multinomial Naïve Bayes and XG Boost for Sentiment Analysis

Social-media provides a vast amount of user-generated content which serves as an invaluable source of real-time data for analyzing public sentiments about various aspects. Public sentiment shapes various services and strategies as per the targeted users. Analyzing text for sentiments or opinions comes under a branch of Natural language processing (NLP), known as Sentiment analysis. Various approaches have been used for sentiment analysis from time to time, viz., machine learning methods, lexicon-based approaches and hybrid of these approaches. This paper uses machine leaning methodologies for sentiment analysis. It explores the effectiveness of Multinomial Naive Bayes (MNB) and Gradient boosting using XG Boost in sentiment analysis. The strength and limitations of both approaches are explored using cross-validation technique. This approach contributes to existing research in the area of sentiment investigation by critically assessing the performance of proposed model and by providing valuable implications in the areas like brand monitoring and public opinion.