Comparative Study: The Implementation of Machine Learning Method for Sentiment Analysis in Social Media. A Recommendation for Future Research

Evolution of Web 2.0 with user generated content produces very large data, is triggered the growth and use of high levels of social media. In the USA only, users of social media since 2006 grew by 356%. Statistics indicate in online 2.27 billion people, 140 million twitter's active users, 50 million active users of Instagram, etc. The phenomenon causes birth-related research and development of social media, one of which sentiment analysis. Research conducted using a variety of methods of sentiment analysis, supervised learning, unsupervised learning (lexicon based) and hybrid methods. This paper is a comparative study on the methods used by the limitations of the use of supervised learning approaches. The comparative study conducted on the study period of 5 years back (2012-2008). The purpose of this study is to make a list of the advantages and disadvantages of the method and data used and need to be improvised, so as to provide recommendations or references for future research.