Classification of financial markets influencers on Twitter

Abstract

Information published by social network users can have an impact on the (virtual and physical) community, which supports the concept of influencer, which is a user generally focused on a set of interests (such as fashion, health, well-being, beauty, finances, etc.), whose opinion correlates with the behavior observed in a community of users who share those interests. The same concept of social network influencer may also be explored in the financial markets, considering user profiles whose expressed opinions align with the performance of a company's shares. Several users interested in the stock exchange use Twitter as a platform to express their opinions on stocks. Considering that tweets express either a positive, negative, or neutral feeling about a stock, we present a model to describe the profile of an influencer, as well as analyze the correlation between their tweets and the performance of a stock. Relying on a genetic algorithm to best describe an influencer’s profile and selecting the best fit from a set of features, this study provides a comparison between sentiment analysis techniques, applied on social network content, produced by potential influencers. We present a study of several sentiment analyses approaches, namely using Support Vector Machines, Neural Networks, Naïve Bayes, and K-Nearest Neighbors. Results show that the method with the best performance is the Neural Network. This outcome is used as a parameter for analyzing an influencer's tweets. Validation showsthat, by integrating a genetic algorithm for influencer detection with a sentiment analysis technique based on Neuronal Networks, it is possible to identify users with social network content sentiment score correlated with variations in a share price, thus representing an information signal suitable to be used by prediction models.