The Performance Evaluation of Machine Learning Classifiers on Financial Microblogging Platforms

As technological advancements facilitate democratization of knowledge, Microblogging platforms are vying to become the premier source of knowledge and are competing with news outlets. A huge number of messages is generated on different microblogging platforms. In financial markets, microblogging websites, such as StockTwits, have become a rich source for amateur investors, which make them ideal sources for market sentiment analysis. Indeed, StockTwit has been widely used by researchers for sentiment analytics and market predictions. However, the quality of the sentiment analysis is highly dependent on the machine learning classifiers used as well as the preprocessing of data. In this study, we compare the performance efficiency of different machine learning classifiers on the user-generated content on StockTwits. We find that Logistic Regression Classifier performs best in a 2-way classification of StockTwits data. Our results report better classification accuracy than a similar research using data from Twitter. We have discussed managerial implications of our results.