Sentiment Analysis of tweets using Supervised Machine Learning

This paper addresses the problem of analyzing sentiment on a social media platform called Twitter; that is to identify and classify whether a tweet expresses a positive sentiment or a negative sentiment. Twitter is an online social networking and micro-blogging website where people from around the world communicate and write short updates, called as “tweets” without exceeding the character limit of 140 characters. It is a continuously expanding web service, having an average of 330 million monthly active users as of the fourth quarter of 2017. This large number of users results in enormous amounts of publicly available data that may be used to gather insight and to reflect the public sentiment by analyzing the tweets. There are many applications to performing sentiment analysis in today’s world in various fields and industries such as gathering market research insights, predicting political campaigns’ outcomes etc. Performing sentiment analysis with traditional techniques of writing manual rules and conditions becomes complex as the number of rules to be written, keep increasing to tackle the raw natural language. The aim of this paper is to investigate further into the sub-domain of natural language processing called sentiment analysis, and develop a statistics based functional classifier and incorporate it with other machine learning models to create a vote based classifier algorithm, for automating and performing accurate sentiment classification of desired tweets.