Credibility assessment of financial stock tweets

Social media plays an important role in facilitating conversations and news dissemination. Specifically, Twitter has recently seen use by investors to facilitate discussions surrounding stock exchange-listed companies. Investors depend on timely, credible information being made available order make well-informed investment decisions, with credibility defined as the believability of information. Much work been done assessing domains such politics natural disaster events, but financial statements is scant within literature. Investments apocryphal could hamper efforts social media’s aim providing a transparent arena for sharing encouraging discussion market events. This paper presents novel methodology assess tweets, which evaluated conducting experiment using tweets pertaining companies listed London Stock Exchange. Three sets traditional machine learning classifiers (using three different feature sets) are trained annotated dataset. We highlight importance considering features specific domain needs be assessed – case this paper, features. In total, after discarding non-informative features, 34 general combined over 15 training classifiers. Results show that both can yield improved performance than alone, Random Forest top performer, although model requires more (37) other (such K-Nearest Neighbours − 9) achieve performance.