Social media are gaining an unprecedented momentum as well as impact on many areas of our life. In this thesis we investigate and analyse the relationship between foreign exchange market (also called Forex) and social media. Specifically, we analyse topics concerning EUR/USD exchange rate because of the large trading volume of the currency pair. We analysed Twitter in the span of three years.
We gathered the data on market movements, events, and Twitter posts. We annotated the tweets with authors' expectation of the Forex movement. By analysing tweets and users tweeting about EUR and USD, we discovered groups of users that behave differently and devised a model for classifying users into these groups. The model is the basis of our research and findings.
We developed a web application for visualisation and browsing the data and results of the analyses. This application enabled fast analysis and debugging of it.
Events connected to EUR and USD have high influence on market movements. We studied the relations between Forex movements and Twitter sentiment around the time of events. We analysed the performance of different user groups around the events (i.e. financial announcements in the USA and EU), which usually result in significant increase of Twitter volume. Predictive performance of the user groups varies in terms of describing market movements. Certain user groups give better results than professional analysts which shows efficiency of our user classification model.
|
