Прогнозирование событий с помощью ленты Twitter

Abstract

The paper explores the possibility of forecasting based on the intelligent analysis of a feed of the Twitter social network. The object of this analysis is the Bitcoin crypto currency market. The theoretical basis for this forecasting approach is the efficient market hypothesis, which states that new information can be used for gaining economic advantage. The source of this information is the sentiment analysis of Twitter messages. The work proposes a new algorithm of determining the emotions of Twitter messages with three possible grade levels: negative, neutral and positive. Thanks to comprehensive correlation analysis and forecasting experiments, it was determined that there is a relation between the analyzed source data, and the prediction accuracy of Bitcoin market movements reached 63,27%.