PREDICTION OF USA ELECTIONS 2020 USING SENTIMENT ANALYSIS

Abstract

Machine Learning (ML) is expanding its applications in our life as the amount of data stored on servers is increasing daily. With ample of provided applications to ease our workload and make us more efficient. exit polls for elections although quite accurate cannot be completely relied upon. This can arise due to pressure from political leaders, peers; people who don't want to share their views etc. There have been instances in past where the results of elections were completely contradictory to predictions based on exit polls. With social media, people have become more vocal about their views and perspectives with the privacy and security over internet. With more people using social media to express their views, we can create several detailed and structured datasets according to our needs. this decreases time as compared to interviewing one person at a time, we can get data of millions promptly. this data can be classified on the basis of region, age, gender, etc. Using ML algorithms on these datasets we can predict the sentiment of these people and can get an accurate prediction for the elections.We'll be performing sentiment analysis on one such dataset which consists of tweets extracted from Twitter. This report will include using seven algorithms: Dictionary based, Naive Bays, Support Vector Machine, Linear Regression, Logistic Regression and Bayesian network and Bayesian network and compare the results of these Algorithms along with their accuracies.