Deep Learning Models for High-Frequency Market Microstructure Data
We present recent work on a large-scale deep learning model to predict price movements from limit order book (LOB) data of cash equities. The model, which utilises convolutional filters and LSTMs, is trained using electronic market quotes from the London Stock Exchange. Our model delivers a remarkably stable out-of-sample prediction accuracy for a variety of instruments and outperforms existing methods. Interestingly, our model translates well to instruments which were not part of the training set, indicating the model's ability to extract universal features. Furthermore, we can extend our framework to incorporate ideas from Bayesian neural networks. Those Bayesian techniques not only deliver uncertainty information that can be used for trading but also improve predictive performance as stochastic regularisers. To the best of our knowledge, we are the first to apply Bayesian networks to microstructure data. (Based on recent work in collaboration with Z. Zhang and S. Roberts, IEEE Trans. Sig. Proc. (u.r.) and NIPS 2018.)
Stefan Zohren is an Associate Professorial Research Fellow at Machine Learning Research Group and the Oxford-Man Institute for Quantitative Finance, University of Oxford. He also acts as a consultant for Man Group. Prior to that, Stefan worked on equities market making as a quant researcher/trader at two leading HFT firms in London and coordinated the Quantum Optimisation and Machine Learning project, a joined research project of Oxford University, Nokia Technologies and Lockheed Martin. His background is in theoretical physics, probability theory and statistics. Stefan's research interests include statistical physics approaches to machine learning, information theory and optimisation, quantum computing as well as machine learning applied to finance, particularly market microstructure and high-frequency data.