Improvements to Multinomial Naïve Bayes for Increasing the Accuracy of Aggressive Tweets Classification

Abstract: Naïve Bayes is a popular supervised learning method widely used for text classification and sentiment analysis. There has been a rise of aggressive troll comments in the social networking sites which leads to online harassment and causes distressful online experiences. This paper uses  Naïve Bayes classifier using Bag of Words on ‘Tweets dataset for Detection of Cyber-Trolls’ (dataset taken from Kaggle) and aims to improve baseline model by adding cumulative changes and studying their impact on the performance of the model.

Keywords: Naïve Bayes, classification, improvements, accuracy.