Combining Likes-Retweet Analysis and Naive Bayes Classifier within Twitter for Sentiment Analysis

Abstract

Sentiment analysis is a research study that aims to extract subjectivity of opinions. Due to massive growth number of user generated content in social media, Twitter is one of the most popular microblogging application which user is freely to discuss and share opinions about specific topic or entity. Twitter have several features that potentially can be used to improve sentiment analysis such as like and retweet. Like and retweet are mechanism in Twitter to propagate or share and to show appreciation of other user posting. This paper proposes a combination of textual and non-textual features to improve performance of sentiment prediction. In this research we apply Naïve Bayes for textual classification and Fisher Score to determine non-textual (like and retweet) features. By combining two kinds of features, our experimental find the optimal value of α and β. The evaluation performance using F1-measure gives 0.838 of accuracy with α and β are 0.6 and 0.4 respectively.