This notebook contains an independently developed Keras/Tensorflow implementation of the CNN-LSTM model for Limit Order Book forecasting originally proposed by Zhang et al. (https://arxiv.org/pdf/1808.03668.pdf). The current implementation was adopted in the paper written by Briola et al.(https://arxiv.org/pdf/2007.07319.pdf).
The interested reder can find the original implementation of the model at https://github.com/zcakhaa/DeepLOB-Deep-Convolutional-Neural-Networks-for-Limit-Order-Books/blob/master/jupyter/run_train_represent.ipynb.
If you use the proposed Keras/Tensorflow implementation in your work please cite:
@article{briola2020deep, title={Deep Learning modeling of Limit Order Book: a comparative perspective}, author={Briola, Antonio and Turiel, Jeremy and Aste, Tomaso}, journal={arXiv preprint arXiv:2007.07319}, year={2020} }
and
@misc{briola_antonio_and_turiel_jeremy_david_2020_4068530, author = {Briola, Antonio and Turiel, Jeremy David}, title = {CNN-LSTM\_Limit\_Order\_Book\_Tensorflow}, month = oct, year = 2020, publisher = {Zenodo}, version = {v1.1}, doi = {10.5281/zenodo.4068530}, url = {https://doi.org/10.5281/zenodo.4068530} }
Please always remember to cite the original papers this work is based on:
@article{zhang2019deeplob, title={Deeplob: Deep convolutional neural networks for limit order books}, author={Zhang, Zihao and Zohren, Stefan and Roberts, Stephen}, journal={IEEE Transactions on Signal Processing}, volume={67}, number={11}, pages={3001--3012}, year={2019}, publisher={IEEE} }