ENSEMBLE BASED TWITTER SENTIMENT ANALYSIS ON BIG DATA USING APACHE SPARK

Abstract

The development of social media platforms like Twitter and Facebook has been a revolution in the history of mankind. Social media web applications played a great role in getting the world together. There are platforms like twitter where people can share their opinion about an issue which can be seen and heard by the entire world. Tweet count on twitter for a day is about 500 million tweets. By analysing these tweets on sentiment basis we can understand the public opinion regarding a particular matter. The system developed will perform a sentiment analysis on live twitter data using an ensemble algorithm to detect the nature of tweets (tweets are positive or negative). Implementing Sentiment analysis on twitter data is quite a difficult task as it involves operating on a large amount of diverse data. This thesis involves the implementation of the apache spark framework for working with a huge chunk of data that is big data. Sentiment analysis is a machine learning-based method. This thesis aims to examine various classification algorithm and their impact on live twitter data.