ChatGPT-based Sentiment Analysis and Risk Prediction in the Bitcoin Market

Abstract

The risk prediction of financial markets is of paramount importance, with investor sentiment playing a critical role. However, current research appears to be lacking in-depth exploration of this particular aspect within the Bitcoin market. This study aims to explore the impact of market participants’ sentiment on risk prediction in the bitcoin market. We first applied ChatGPT to analyze the sentiment of crawled Bitcoin-related news headlines. Meanwhile, Monte Carlo simulation was employed to calculate value at risk (VaR). And we selected five conventional factors, including Bitcoin price, transaction volume, market share, hash rate, and average difficulty of mining. Finally, K-Nearest Neighbors (KNN) regression model was used to construct the model for predicting the risk of bitcoin market. We made a comparison between the accuracy outcomes when considering and not considering sentiment as factors. The results show that market participant’s sentiment is significantly associated with market risk, and the inclusion of sentiment can significantly improve the accuracy of the risk prediction model.