Abstract
This paper explores multidisciplinary application of emotions and human psychology within market behavior to analyse how emotional dimensions, extracted from Twitter messages and related to cryptocurrency market, are connected to future uncertainty and risk exposure. Although Twitter messages are often used to derive sentiment scores that are then linked to market performance, specific values of emotional components have not been utilised in the previous academic literature to analyze risk behavior. We connect VAD (valence, arousal and dominance) dimensions with the future hourly absolute value of returns, future 24-hour return standard deviation, future 24-hour downside deviation and future 24-hour maximal drawdown (all measuring market uncertainty and risk) on the cryptocurrency market. Results show that all target variables have various predictability from VAD dimensions, where lagged dominance variable (perceived level of control) is the key driver of predictability. Additionally, VAD dimensions have been grouped into distinct clusters by using the k-Means approach. A comparison of selected clusters on in-sample and out-of-sample data showed consistent predictability of identified clusters to all target variables. Results show that emotional components of human emotions, derived from cumulative Twitter messages, actually predict future uncertainty and risk with consistent clustering profile from lagged dominance VAD dimension where lower dominance values predict higher future risk and vice versa.
Highlights
Textual NLP analysis has been a fruitful area of research in the last decade, analysing everything from long books, news article and all the way to short Twitter messages
Since most sentiment analysis operate on the level of valence emotional scale our results show previously undiscovered usefulness of dominance dimension for risk assessment and prediction
We show that average VAD dimensions from words on hourly level have a predictable component that is able to predict all tested future uncertainty and risk target variables
Summary
Textual NLP analysis has been a fruitful area of research in the last decade, analysing everything from long books, news article and all the way to short Twitter messages. Since emotions influence human perception of the situation, and inherently human behavior, multiple studies have come up to find connections between sentiment and market behavior [6]. Through machine learning this has resulted in multiple analysis of both news sentiment and Twitter sentiment trying to predict both market returns and market risk. Risk component is key analysis component of this paper where several approaches exist to measure both uncertainty and risk in the underlying market behavior. RQ2 - is there connection between VAD dimensions and future 24 hour standard deviation of market returns,. RQ3 - is there connection between VAD dimensions and future 24 hour downside deviation of market returns,. This paper is organized as follows: section ‘‘Related work’’ analyses in detail components from the research questions, section ‘‘Data and methodological approach’’ gives overview of analysed data and testing approach, section ‘‘Empirical findings’’ presents empirical findings and results, section ‘‘Discussion and results summary’’ analyses found results and presents them in a coherent manner, and section ‘‘Conclusion’’ summarises relevant discoveries with an emphasis on key conclusions and future possibilities
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
