The current project page provides pytorch code that implements the following paper:
Title: "Benchmark Dataset for Short-Term Market Prediction of Limit Order Book in China Markets"
Authors: Charles Huang, Weifeng Ge, Hongsong Chou, Xin Du
Institution: FinAI Laboratory, Hong Kong Graduate School of Advanced Studies
Code: https://github.com/hkgsas/LOB/tree/master/lob_modeling
Paper: https://github.com/hkgsas/LOB/blob/master/Benchmark%20Dataset%20for%20Short-Term%20Market%20Prediction%20of%20Limit%20Order%20Book%20in%20China%20Markets%202020%20Nov%20v3.pdf
Abstract: Limit Order Book (LOB) has generated “big financial data” for analysis and prediction from both academic community and industry practitioners. This paper presents a benchmark LOB dataset of China stock market, covering a few thousand stocks for the period of June to September 2020. Experiment protocols are designed for model performance evaluation: at the end of every second, to forecast the upcoming volume-weighted average price (VWAP) change and volume over 12 horizons ranging from 1 second to 300 seconds. Results based on linear regression model and state-of-the-art deep learning models are compared. Practical short-term trading strategy framework based on the alpha signal generated is presented.
This code was developed and tested with pytorch version 1.7.0
You can download lob dataset from here.
Download the dataset, and put them in the path 'tt.arg.dataset_root/'
.\LOB_data
.\2020.6
.\2020.7
.\2020.8
.\2020.9
lob_sz_6789_train_val.txt
lob_sz_678_train.txt
lob_sz_9_val.txt
In train.py, replace the dataset root directory with your own:
tt.arg.dataset_root = './LOB_data/'
# ************************** linear regression *****************************
$ CUDA_VISIBLE_DEVICES=0 python3 train.py --model linear_regression --device 0 2>&1 | tee loss.log
# ************************ multilayer perceptron ***************************
$ CUDA_VISIBLE_DEVICES=0 python3 train.py --model mlp --device 0 2>&1 | tee loss.log
# ******************************** LSTM ************************************
$ CUDA_VISIBLE_DEVICES=0 python3 train.py --model lstm --device 0 2>&1 | tee loss.log
# ********************************* CNN ************************************
$ CUDA_VISIBLE_DEVICES=0 python3 train.py --model cnn --device 0 2>&1 | tee loss.log
# ****************************** CNN-LSTM **********************************
$ CUDA_VISIBLE_DEVICES=0 python3 train.py --model cnn_lstm --device 0 2>&1 | tee loss.log
Please set these parameters accordingly.
tt.arg.model = 'cnn_lstm' if tt.arg.model is None else tt.arg.model
tt.arg.input_size = 124 if tt.arg.input_size is None else tt.arg.input_size
tt.arg.class_num = 5 if tt.arg.class_num is None else tt.arg.class_num
tt.arg.epoch = 20 if tt.arg.epoch is None else tt.arg.epoch
tt.arg.timestep = 3 if tt.arg.timestep is None else tt.arg.timestep
tt.arg.sequence_len = 50 if tt.arg.sequence_len is None else tt.arg.sequence_len
tt.arg.model_type = None if tt.arg.model_type is None else tt.arg.model_type