Sentiment Analysis Using ML Classifiers and Recurrent Neural Network

Abstract

Sentiment Analysis is an efficient tool businesses use to understand the sentiment their brand has among customers. Using Machine Learning and Neural Networks we can automate this process for businesses and enable them to identify customer perspective from tens of thousands of texts in a short time. In our project, we performed sentiment analysis on a twitter dataset and classified the tweets on basis of their polarity, positive or negative. To achieve this we used 4 Machine Learning classifiers: Logistic Regression, Linear SVM, Naïve Bayes and Random Forest. Additionally, we also used a recurrent neural network. The results of all the classifiers were compared and it was found that the RNN outperforms the ML classifiers by a major margin. Among the ML classifiers, Logistic Regression was found to give the best results although all of the accuracy of all these classifiers was close to each other.