Comparison of Machine Learning Techniques for Sentiment Analysis
Shashank Mishra, Mukul Aggarwal, Shivam Yadav, Yashika Sharma
Abstract
Sentiment analysis is the process of categorizing and locating the emotions represented in a textual source. Sentiment analysis can be used widely in different areas, such as customer review data, feedback data classification, survey responses, and social media comments. Tweets on Twitter contain a variety of sentiments reflecting the perception, thinking, and working background of the user. With the help of the sentiment analyzer, it can define the response of others on any matter or subject of interest. Here, we used machine learning-based NLP (natural language processing) and text analysis technology to define an automated model that can classify the sentiment of a large dataset. Here we used the following three machine-learning classifiers: logistic regression, SVM, and Bernoulli Naïve Bayes. The effectiveness and performance of these classifiers are assessed using F1 scores and accuracy. The accuracy of these models is 83%(LR), 81%(SVM), and 80%(BNB) So logistic regression model provides the best result among these.