Sentiment prediction using enhanced xgboost and tailored random forest

Abstract

A large quantity of data is being generated in the form of blogs, tweets and updates of opinions on the topic of interest. People give their feelings and opinions on different topics such as movies, products, education, politics, news and so on. Analysis of such data is very useful to understand the views/opinions/sentiments of the society. Such analysis would also be more useful in decision making . The major challenge in analysis is the usage of jorgon words, spelling mistakes, hash tags, hyperlinks and irrelevant words. This research aims to know the opinion of people on particular topics considering their tweets. These can be evaluated as classification problem to analyse the tweets expressed in texts for hidden sentiments. For this purpose, we proposed and evaluated a tailored random forest and enhanced XGBoost algorithms. We achieved significantly better accuracy by enhancing XGBoost compared to tailored random forest and naive bayes for tweets classification.