Bitcoin returns and volume: Uncovering investor sentiment through topic modeling
本文利用主题建模分析Bitcointalk.org上的讨论，发现八个比特币相关主题，并揭示主题特定的情绪、分歧和关注度与比特币回报及交易量之间的关系。
This paper examines discussions on Bitcointalk.org . Using topic modeling analysis, we discover eight distinct bitcoin-focused topics: Investment , Public Perception , Regulation , Wallet , Bitcoin's Nature , Payment , Mining , and Other . We uncover variations in how topic-specific sentiment, disagreement, and attention relate to bitcoin returns and trading volume. Specifically, among all topics, only the sentiment and disagreement of the Investment topic exhibit a significant relation with bitcoin returns. However, the sentiment, disagreement, or attention of several topics, such as Regulation and Wallet , shows a significant relation with bitcoin returns only at the extremes of the return distribution (during up and down markets). In contrast, the sentiment, disagreement, and attention of each topic have a significant association with bitcoin trading volume across the entire distribution. Importantly, topic-level sentiment and attention exhibit asymmetric relations with bitcoin returns.