The role of Guru investor in Bitcoin: Evidence from Kolmogorov-Arnold Networks

This study examines the influence of Twitter sentiment on Bitcoin price movements by distinguishing between "Gurus" (influential users) and regular users in the Bitcoin market. We analyze over 26 million Tweets collected from September 2006 to March 2023 to derive sentiment data, then employ Kolmogorov-Arnold Networks (KAN) to compare the predictive effectiveness of follower-weighted sentiment versus unweighted sentiment. Our results indicate that follower-weighted sentiment significantly enhances prediction accuracy, with Guru sentiments consistently showing stronger predictive power than regular user sentiment. These findings are robust to alternative measurement of sentiment, alternative definition of Guru investor, and subperiod analysis.