Sentiment Analysis of Social Media Data using Fuzzy-Rough Set Classifier for the Prediction of the Presidential Election

As an interdisciplinary research field, sentiment analysis is one of the momentous applications in Natural Language Processing, for quantifying the emotional value in vast data in the form of text available in social media networks to gain an understanding of the attitudes, opinions, and emotions expressed. There is a great deal of literature on the various approaches to address sentiment analysis with social media and this research focuses on Machine Learning techniques with Twitter data analysis. Special attention is drawn towards the classifiers based on the Fuzzy Set and Rough Set approach which are two powerful mathematical components of computational intelligence with its new dimension involved in the field of sentiment analysis. However, there is a minimal number of review papers discussing rough-fuzzy classifier involvement in sentiment analysis and there is a plethora of work that must be done with text mining in natural language processing. The mission of this study is to develop a sentiment-based classifier using machine learning and fuzzy-rough set theory. Further, it carries automatic sentiment classification with Twitter corpus collected during September 1^st and November 15^th, 2019 (two months before the election) regarding the case study for the prediction of results at the presidential election 2019, Sri Lanka. The fuzzy rough classifier is developed using the Fuzzy Rough Nearest Neighbor algorithm. The accuracy of the fuzzy rough set-based classifier is higher compared to other classifiers. The actual results of the presidential election of 2019 are tally with the predicted results of the classifier. Therefore, the current state of the art for the prediction of political sentiment with microblogging is probable with the social media data as witnessed with this case study and this can be used in other cases as well.