基於限價委託簿，使用深度學習模型預測比特幣期貨之未來價格漲跌

Predicting Bitcoin Futures Price Movements by Deep Learning Based on the Limit Order Book

陳韋勳(Wei-Shiun Chen)

指導教授 ： 呂育道

國立臺灣大學/管理學院/財務金融學系/碩士(2023年)

https://doi.org/10.6342/NTU202302237

摘要

近年來比特幣期貨市場發展迅速，成為投資人熱門的領域之一，本研究採用經過傳統金融市場驗證的深度學習模型DeepLOB，利用幣安交易所的比特幣期貨限價委託簿預測該市場未來價格的漲跌。透過與傳統線性模型的比較，發現DeepLOB模型具有更準確的預測能力，顯示其在擷取市場特徵和捕捉趨勢方面表現得更優異。此外，由於DeepLOB模型使用的限價委託簿表示法會導致其不穩定的預測結果，本研究採用Wu、Mahfouz、Magazzeni和Veloso提出的限價委託簿穩健表示法來取代模型原始的方法，實驗結果顯示新的表示法有助於提升DeepLOB模型的預測能力。
總體而言，本研究顯示DeepLOB模型在幣安交易所的比特幣期貨市場中比傳統線性模型表現得更優異，且透過使用限價委託簿穩健表示法可以進一步提升預測準確度。這些結果為投資人提供有益的參考，協助他們在比特幣期貨市場中做出更明智的決策。

關鍵字

比特幣期貨 ； 幣安交易所 ； 限價委託簿 ； DeepLOB ； 穩健表示法

並列摘要

In recent years, the Bitcoin futures exchanges have experienced rapid development and have become a popular area for investors. This thesis applies the DeepLOB, a deep learning model validated in traditional financial markets. It utilizes the limit order book of Bitcoin futures from the Binance exchange to forecast futures price movements. Through a comparison with traditional linear models, it is found that the DeepLOB model demonstrates more accurate predictive capabilities, indicating its superior performance in capturing market features and trends.
The original representation of the limit order book in the DeepLOB model can lead to unstable predictions. This thesis adopts the robust representation of Wu, Mahfouz, Magazzeni and Veloso (2021). Experimental results show that the new representation contributes to improving the predictive capabilities of the DeepLOB model.
Overall, this thesis demonstrates that the DeepLOB model outperforms traditional linear models for the Bitcoin futures traded on the Binance exchange, and the robust limit order book representation further enhances accuracy in prediction. These findings provide valuable insights for investors, assisting them in making more informed decisions in the Bitcoin futures markets.

並列關鍵字

Bitcoin futures ； Binance Exchange ； limit order book ； DeepLOB ； robust representation

參考文獻

Foucault, T., Kadan, O., & Kandel, E. (2005). Limit order book as a market for liquidity. Review of Financial Studies, 18(4), 1171–1217.

Géron, A. (2022). Hands-on Machine Learning with Scikit-Learn, Keras, and TensorFlow. Sebastopol, CA: O'Reilly Media.

Graves, A. (2012). Long Short-Term Memory. Berlin, GER: Springer.

Hung, J. C., Liu, H. C., & Yang, J. J. (2021). Trading activity and price discovery in Bitcoin futures markets. Journal of Empirical Finance, 62, 107–120.

Kleinbaum, D. G., Dietz, K., Gail, M., Klein, M., & Klein, M. (2002). Logistic Regression. New York: Springer-Verlag.