An Efficient deep learning model to Predict Stock Price Movement Based on Limit Order Book

TL;DR

This work tackles the challenge of predicting stock price movements from high-dimensional Limit Order Book (LOB) data in high-frequency trading by exploiting symmetry between the bid and ask sides. It introduces a Siamese architecture with shared encoders to process ask and bid data separately, complemented by a decoder that uses the difference between the two representations, and proves its effectiveness across multiple baselines using OFI features on 14 Chinese A-share stocks. The study also investigates the role of Multi-Head Attention (MHA) when combined with LSTM, finding that MHA can boost short-horizon forecasts, especially with strong input features. Overall, the proposed symmetry-aware Siamese framework consistently improves predictive performance over a range of models and inputs, offering a general and transferable approach for LOB-based price forecasting in markets such as the Chinese A-share market, with potential applicability to other market settings and future trading strategies.

Abstract

In high-frequency trading (HFT), leveraging limit order books (LOB) to model stock price movements is crucial for achieving profitable outcomes. However, this task is challenging due to the high-dimensional and volatile nature of the original data. Even recent deep learning models often struggle to capture price movement patterns effectively, particularly without well-designed features. We observed that raw LOB data exhibits inherent symmetry between the ask and bid sides, and the bid-ask differences demonstrate greater stability and lower complexity compared to the original data. Building on this insight, we propose a novel approach in which leverages the Siamese architecture to enhance the performance of existing deep learning models. The core idea involves processing the ask and bid sides separately using the same module with shared parameters. We applied our Siamese-based methods to several widely used strong baselines and validated their effectiveness using data from 14 military industry stocks in the Chinese A-share market. Furthermore, we integrated multi-head attention (MHA) mechanisms with the Long Short-Term Memory (LSTM) module to investigate its role in modeling stock price movements. Our experiments used raw data and widely used Order Flow Imbalance (OFI) features as input with some strong baseline models. The results show that our method improves the performance of strong baselines in over 75$% of cases, excluding the Multi-Layer Perception (MLP) baseline, which performed poorly and is not considered practical. Furthermore, we found that Multi-Head Attention can enhance model performance, particularly over shorter forecasting horizons.

Figures (10)

• Figure 1: The limit order book data structure.
• Figure 2: The train/valid/test sets split.
• Figure 3: The CNN-LSTM model architecture overview.
• Figure 4: The LSTM-MHA model architecture overview.
• Figure 5: The Siamese network architecture.