Online Hybrid Neural Network for Stock Price Prediction: A Case Study of High-Frequency Stock Trading in the Chinese Market

Time-series data, which exhibit a low signal-to-noise ratio, non-stationarity, and non-linearity, are commonly seen in high-frequency stock trading, where the objective is to increase likelihood of profit by taking advantage tiny discrepancies prices trading on them quickly huge quantities. For this purpose, it essential apply method that capable fast accurate prediction from such time-series data. In paper, we developed an online time series forecasting for (HFT) integrating three neural network deep learning models, i.e., long short-term memory (LSTM), gated recurrent unit (GRU), transformer; abbreviate new LGT or O-LGT. The key innovation underlying our its efficient storage management, enables super-fast computing. Specifically, when computing forecast immediate future, only use output calculated previous data (rather than themselves) together with current Thus, involves updating into process. We evaluated performance O-LGT analyzing limit order book (LOB) Chinese market. It shows that, most cases, model achieves similar speed much higher accuracy conventional supervised models HFT. However, slight sacrifice accuracy, approximately 12 64 times faster existing high-accuracy LOB