Sentiment Analysis System for Classification of Patient-Generated Health Reviews Using Rough Set Theory and Machine Learning Technique

Abstract

Advancements in social media domain have led to a prominent progress in the number of online communities. Sites, such as Twitter and Facebook, provide an avenue for the unrestricted generation, communication, and distribution of messages as well as information. In this work, we propose a sentiment classification system from patient-generated content posted by users on medical forums and social media sites. The rough set theory is a numerical rule-based technique employed for categorizing and examining doubtful, partial or indistinct data. The emphasis of this study is on the employment of the rough set theory technique for sentiment classification of patient-generated health reviews. We investigated four rough set theory-based algorithms, namely: Genetic, Learning from Examples Module version 2 (LEM2), Exhaustive and Covering, to generate rules for sentiment classification of patient-generated health reviews text. The Rough Set Exploration System (RSES 2.0) software is utilized to conduct experiments. Additionally, we applied SVM classifier to classify emotions. The experimental results show that the Genetic algorithm outperforms the comparing algorithms with an accuracy of 84.2% and Support Vector Machine outperforms other classifiers with an accuracy of 80.5%.