jonchin / temporal_relational_stock_ranking Goto Github PK

Code for the Relational Stock Ranking (RSR) model and the Temporal Graph Convolution in our paper "Temporal Relational Ranking for Stock Prediction", [paper].

Python 3.6 & Tensorflow > 1.3

All data, including Sequential Data, Industry Relation, and Wiki Relation, are under the data folder.

Raw data: files under the google_finance folder are the historical (30 years) End-of-day data (i.e., open, high, low, close prices and trading volume) of more than 8,000 stocks traded in US stock market collected from Google Finance.

Processed data: 2013-01-01 is the dataset used to conducted experiments in our paper.

To get the relation data, run the following command:

tar zxvf relation.tar.gz

Under the sector_industry folder, there are row relation file and binary encoding file (.npy) storing the industry relations between stocks in NASDAQ and NYSE.

Under the wikidata folder, there are row relation file and binary encoding file (.npy) storing the Wiki relations between stocks in NASDAQ and NYSE.

To repeat the experiment, i.e., train a RSR model, downloaded the pretrained sequential embedding, and extract the file into the data folder.

python relation_rank_lstm.py -rn wikidata -l 16 -u 64 -a 0.1

python relation_rank_lstm.py -m NYSE -l 8 -u 32 -a 10 -e NYSE_rank_lstm_seq-8_unit-32_0.csv.npy

to enable gpu acceleration, add the flag of:

-g 1

If you use the code, please kindly cite the following paper:

@article{feng2019temporal,
  title={Temporal relational ranking for stock prediction},
  author={Feng, Fuli and He, Xiangnan and Wang, Xiang and Luo, Cheng and Liu, Yiqun and Chua, Tat-Seng},
  journal={ACM Transactions on Information Systems (TOIS)},
  volume={37},
  number={2},
  pages={27},
  year={2019},
  publisher={ACM}
}

[email protected]