Deep learning-based cryptocurrency sentiment construction
- PDF / 3,383,665 Bytes
- 29 Pages / 439.37 x 666.142 pts Page_size
- 47 Downloads / 220 Views
Deep learning‑based cryptocurrency sentiment construction Sergey Nasekin1 · Cathy Yi‑Hsuan Chen2 Received: 9 December 2018 / Accepted: 15 February 2020 © Springer Nature Switzerland AG 2020
Abstract We study investor sentiment on a non-classical asset such as cryptocurrency using machine learning methods. We account for context-specific information and word similarity using efficient language modeling tools such as construction of featurized word representations (embeddings) and recursive neural networks. We apply these tools for sentence-level sentiment classification and sentiment index construction. This analysis is performed on a novel dataset of 1220K messages related to 425 cryptocurrencies posted on a microblogging platform StockTwits during the period between March 2013 and May 2018. Both in- and out-of-sample predictive regressions are run to test significance of the constructed sentiment index variables. We find that the constructed sentiment indices are informative regarding returns and volatility predictability of the cryptocurrency market index. Keywords Sentiment analysis · Lexicon · Social media · Word embedding · Deep learning · RNN JEL Classification G41 · G4 · G12
1 Introduction Classical asset pricing theories, mainly relying on the concept of arbitrage, face challenges in the face of a surge of new asset classes like cryptocurrencies. Compared to classical financial assets whose fundamental value can be determined from cash flows such as dividends and earnings, the fundamental value of * Sergey Nasekin [email protected] Cathy Yi‑Hsuan Chen CathyYi‑[email protected] 1
Deutsche Bank AG, Frankfurt, Germany
2
Adam Smith Business School, University of Glasgow, Glasgow, UK
13
Vol.:(0123456789)
Digital Finance
cryptocurrencies is harder to grasp. The techniques behind cryptocurrencies, such as blockchain, ICOs (initial coin offerings), decentralized schemes, complicate fair price estimation. Sentiment plays a significant role in price evolution, given possible arbitrage opportunities and “intangible” fundamental values, see Aboody et al. (2018). Therefore, reliable measurement of cryptocurrency sentiment is particularly of interest. Sentiment provides explanatory power on firms’ future performance, especially when fundamental information is incomplete or biased, see Tetlock et al. (2008), Lerman and Livnat (2010), Feldman et al. (2010) and Loughran and McDonald (2011). There has been a growing body of research recently on predictability of markets’ dynamics using information distilled from textual data sources. These include studies on traditional asset markets’ prediction, see Yu (2013), Plakandaras et al. (2015), Nassirtoussi et al. (2015), Persio and Honchar (2016) as well as cryptocurrencies, see Mai et al. (2018) and Cheuque Cerda and Reutter (2019). News about cryptocurrencies, similar to financial news about stock markets, can be used to construct sentiment. We use a body of text messages from social media used by the crypto community to collect representative
Data Loading...
