A Hybrid Approach to Predict Election Candidate Success Using Candidate Speech and Voter Opinion

Abstract

Election candidate success is a prediction of the winning rate of the candidate. Sentiment analysis on user's social media data plays a prominent role in prediction. It refers to a classification problem where the main goal is to classify data into positive and negative sentiments. Sentiment analysis over user's Twitter data offers an effective way to measure voter opinion towards the candidate. As election forecasting based on the only opinion of the voter is difficult, the proposed system come up with a hybrid approach in which sentiment analysis on user's Twitter data and personality prediction on candidate's speech data is performed. The system highlights the performance of various classifiers. Experimental results show that the logistic regression classifier outperformed the other classifiers.