A comparative study on opinion mining algorithms of social media statuses

Abstract

The Social Media (SM) is affecting clients' preferences by modeling their thoughts, attitudes, opinions, views and public mood. Observing the SM activities is a decent approach to measure clients' loyalty, keeping a track on their opinion towards products preferences or social event. Opinion Mining (OM) is the most rising research field of text mining using Machine Learning (ML) algorithms and Natural Language Processing (NLP). Several algorithms such as Support Vector Machines (SVM), Naïve Bayes (NB) and Maximum Entropy (ME), were utilized to extract information that differentiates the user's opinion whether it's positive, negative or neutral. User's opinions and reviews are very beneficial information for individuals, businesses, and governments. In this paper, we compare the intelligent algorithms, which are utilized for OM in SM data over the last five years. The results show that using SVM with Part Of Speech (POS) or POS, Unigram and Bigram with J48 accomplish Sentiment Classification (SC) accuracy 92%.