Cryptocurrency explainable deep learning price discovery

• In Event: OPERATIONS FINANCE INTERFACE: Application of AI in Financial Services

Abstract

Cryptocurrency and blockchain technologies are gaining significance in the financial market. However, understanding the price dynamics of cryptocurrencies is a challenge in the field of Fintech. To address this issue, we present F-LSTM, a new explainable deep learning model, and a novel feature importance ranking algorithm. Our research focuses on predicting the price trends of the top five traded cryptocurrencies. The results indicate that our proposed model outperforms other peers in forecasting cryptocurrency prices. Additionally, we introduce a relative entropy way to identify the optimal prediction times in deep learning models. It also helps decipher the relationships between different cryptocurrencies

Authors

• Henry Han, Baylor University

• Qiannong (Chan) Gu, Ball State University