As the trend in financial industries to include alternative data into investment decision making continues, social sentiment analysis has become an active field of research. The challenge is to quantify public announcements which have significant impact on financial assets in order to predict price changes. In this study a generic technique, that combines sentiment analysis with short-text topic modelling called Sentiment Biterm Topic Model (sBTM) is proposed. sBTM expands mood time series with topic dimensions. It performs sentiment analysis on latent topic portions, extracted by the Biterm Topic Model, which is specifically designed to classify shorttexts such as tweets. A long-ranging collection of tweets is leveraged to investigate if Twitter activities influence cryptocurrency price formation. Multiple baseline forecasts are explored and finally it is tested whether topic-specific analysis can enhance prediction accuracy. It is shown that the topic based approach is more effective than its non-topic counterpart.
Masterprüfung mit Defensio, Tretzmüller Markus
27.11.2019 09:00 - 10:00
Organiser:
SPL 5
Location:
Besprechungsraum 4.34
Währinger Straße 29
1090 Wien
1090 Wien