Sentiment Analysis Based On Machine Learning For Twitter Accounts

Abstract

Analysis of sentiments on Twitter has become a hot topic of research in recent years. Most of the existing twitter sentiment analysis solution basically only considers Twitter message textual information, and struggles to perform well when confronted with short and ambiguous Twitter messages.  Recent studies show that sentiment diffusion patterns on Twitter are closely related to feeling polarities of Twitter messages.  Therefore, in this paper we concentrate on how to fuse Twitter’s textual messages and feeling diffusion trends in order to achieve better output of Twitter’s sentiment analysis on info.  To this end, we first investigate the diffusion of sentiments by studying a phenomenon called reversal of sentiment, and consider some interesting properties of reversal of sentiment. Then we look at the interrelationships between Twitter’s textual information and sentiment diffusion patterns, and propose an iterative algorithm called SentiDiff to predict the polarities of sentiment expressed in Twitter posts.  To the best of our knowledge, this study is the first to use patterns to feeling diffusion to help develop an interpretation of twitter’s feelings. Extensive experiments on real-world data sets show that, compared to state-of-the-art textual information- based sentiment analysis algorithms, our proposed algorithm yields PR-AUC sentiment-based analysis, our proposed algorithm yields PR-AUC improvements on Twitter sentiment classification tasks between 5:09% and 8.38 %.