The emergence of cryptocurrency markets has drastically changed how online transactions are conducted and provide a new investment opportunity. This study contributes to the literature on directional cryptocurrency price returns prediction by expanding the set of meaningful features extracted from textual data with sentiment analysis and comparing their usefulness across multiple data sources. In contrast to previous studies, we use fine-grained topic-sentiment features. More specifically, aspect-based sentiment analysis models, JST and TS-LDA, are implemented to incorporate joint topical-sentiment features and the degree of text subjectivity. We collected, and make available, a dataset, which consists of data scraped from Reddit, Bitcointalk and CryptoCompare sources, to demonstrate that proposed features lead to interpretable topics and an improvement in predictive performance.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.