Multinomial Naive Bayes Classification Model for Sentiment Analysis

Automatic document sorting becomes increasingly important as handling and organizing documents manually is a time consuming and not a viable solution on given the number of documents is very huge. The Naive Bayes method is very well-known method for text classification due to its effective grating assumptions, quick and easy implantation. In this article, we propose the simple, heuristic solutions to some problems with multinomial Naive Bayes (MNB) that address both systemic problems and those problems that arise due to reason that text is not actually the case generated according to a multinomial model. An MNB classifier is a type of NB classifier and is often used as a baseline for text classification but here it is applied for Sentiment Analysis (SA). We have used a dataset of movie reviews from the site. In each review contains a notice in the form of text and a numerical score (0 to 100 scale). The Exhaustive experiments with a large number of widely used reference data sets for text classification confirm the effectiveness of our proposed algorithm. Thus, accuracy can be greatly improved with Multinomial Naive Bayes classifier.