Comparison between Multinomial and Bernoulli Naïve Bayes for Text Classification

Document/Text Classification has become an important area in the field of Machine Learning. On account of its wide applications in business, ham/spam filtering, health, e-commerce, social media sentiment, product sentiment among customers etc., various approaches have been devised to accurately predict the category or to classify any of the new text/document under consideration. Nowadays, news articles in the newspaper present various kinds of sentiments or inclination of the news article towards a negative or positive sentiment and hence, the content of the news can actively be used to judge the impact on the reader. The paper aims to predict that whether the sentiment of the news article is positive or negative using the two popular approaches of Naïve Bayes Text Categorization i.e. Multivariate Bernoulli Naïve Bayes Classification and Multinomial Naïve Bayes Classification. Also, the research aims to identify that which approach between the given two approaches perform better for the given dataset.