Exchange Rate Movements Prediction Based on Behavioral Analysis of Social Networks Data

In this study, we carried out behavioral Twitter data analysis by semantic analysis of text and identify tone of messages. Two systems have been implemented: a system based on the lexical approach to sentiment analysis, and the second - on machine learning with a teacher. The paper describes an approach to data discovery, ways of presenting text for different machine learning classification algorithms and performance evaluation metrics. A comparison of the work machine learning algorithms classify messages from Twitter in the presentation of the feature vector as the uni-gramm and bi-grams. In the developed system of data processing results was tested the hypothesis predictive potential of Twitter users sentiments for the exchange rate USD / RUB. It was found between the volatility of the exchange rate by the user and thus concluded that the analysis of user behavior in the Twitter can be used to make investment decisions.