Hyperbolic Online Time Stream Modeling

This repository contains the code for Hyperbolic Online Time Stream Modeling accepted at SIGIR 2021 and implements the hyperbolic time aware LSTM as well as extends it to a hierarchical hyperbolic RNN model.

Dependencies:

1. PyTorch
2. geoopt==0.1.2

Please follow the FAST repository to obtain the data. This code expects data in a pickle format, with separate files for training and testing (dummy data to be uploaded soon!).

To train a model for stock movement prediction or profitability, from "code/", run:

python -W ignore train.py --task movement --data stock(or china) --lr 5e-4 --num_epochs 500 --decay 1e-5 --batch_size 128 --name exp_name

To train a model for stock volatility prediction or profitability, from "code/", run:

python -W ignore train.py --task volatility --data stock(or china) --lr 5e-4 --num_epochs 500 --decay 1e-5 --batch_size 128 --name exp_name

To calculate the profitability, from "code/", run:

python stock_trade.py --model_path /path/to/model.pth --data stock(or china)

Acknolwedgements:

1. Hyrnn code: https://github.com/ferrine/hyrnn
2. Manifolds and RAdam optimizer: https://github.com/HazyResearch/hgcn

If you use our models, consider citing:

@inproceedings{10.1145/3404835.3463119,
author = {Sawhney, Ramit and Agarwal, Shivam and Thakkar, Megh and Wadhwa, Arnav and Shah, Rajiv Ratn},
title = {Hyperbolic Online Time Stream Modeling},
year = {2021},
isbn = {9781450380379},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3404835.3463119},
doi = {10.1145/3404835.3463119},
booktitle = {Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval},
pages = {1682–1686},
numpages = {5},
keywords = {finance, language processing, hyperbolic geometry, stock market},
location = {Virtual Event, Canada},
series = {SIGIR '21}
}