The authors of this research paper have developed a deep learning model that can predict price movements in cash equities from limit order book (LOB) data. The architecture of the model includes convolutional filters to capture the spatial structure of LOBs and LSTM modules to capture longer time dependencies.
The proposed network outperforms all existing state-of-the-art algorithms on the benchmark LOB dataset. In a more realistic setting, the model is tested using one year market quotes from the London Stock Exchange and delivers a stable out-of-sample prediction accuracy for a variety of instruments.
Additionally, the model translates well to instruments that were not part of the training set, indicating the model’s ability to extract universal features. The authors also perform a sensitivity analysis to understand the rationale behind the model predictions and reveal the components of LOBs that are most relevant. The ability to extract robust features that translate well to other instruments is an important property of the model and has many other applications.
Summary: This research paper introduces a hybrid deep neural network that can predict stock price movements using high frequency limit order data. The proposed method is evaluated against several baseline methods on the FI-2010 benchmark dataset and the results show that it performs better than other techniques in predicting short-term price movements. The model is also tested using one year of limit order data from the London Stock Exchange and it is observed that the model generalizes well to instruments that were not part of the training data. A simple trading simulation is also used to test the model, yielding good profits that are statistically significant. The authors also use a method for sensitivity analysis to indicate the components of inputs that contribute to predictions and understand the relationship between the input’s components and the model’s prediction.
— 2023年1月24日