Machine Learning Based Emotional Sentiment Analysis of Tweet Data Using a Voting Classifier

Abstract

The introduction of social media and microblogging sites to the World Wide Web was a significant advancement. The website functioned as a place for users to express their opinions and feelings on a variety of problems. The Internet has bloomed into a viable platform for online education, information distribution, and the expression of varied opinions since the dawn of the social networking age. Social media networks contain a multitude of sentiment data in the form of tweets, blogs, status updates, articles, and so on. This study takes advantage of Twitter, the most popular microblogging network. Sentiment analysis is used to derive user views and sentiments from Twitter data (tweets). Some of the seven machine learning models presently employed by the existing system to categorise tweets into happy or sad categories include SVM, DTC, NB, RF, GBM, LR, VC (LR+SGD), and VC (LR+SGD). Following a thorough performance comparison, it was discovered that the voting classifier (LR-SGD) in conjunction with the topic-based information content index (TF-IDF) produces the best results, with an F1 score of 81% and an accuracy of 79%. The suggested system consists of combining the LR, RF, NB, and SVM voting classifiers with the TF-IDF model, which produces 94% accuracy, and the COUNT vectorization model, which yields 95% accuracy. The findings might help governments and companies improve the execution of programmes, goods, and events.