Stock Movement Prediction Based on Bi-typed and Hybrid-relational Market Knowledge Graph via Dual Attention Networks

Requirement Environment

• Python 3.6.13
• Pytorch 1.7.1
• Geometric 1.7.2

Run

$ python main1.py

• Make sure that the GPU is used to reproduce our experiments.

Data

The two datasets for SMP with their folder names are given below.

• CSI100E
• CSI300E.

Selected Stock

• The selected stocks in CSI100E and CSI300E can be found at ./raw_data/100E/stocks.txt and ./raw_data/100E/stocks.txt,respectively

Transcational Data

• Due to The raw transcational data are about 1.1Gb and the limited space, a small part of the data is given. more data and preprocess code will be released soon.

• A part of transcational data can be found at ./raw_data/100E/historical price.xlsx and ./raw_data/300E/historical price.xlsx.

Sentiment Indicators

• Due to the limited space and news data are about 2.3Gb, a small part of the data is given. more data and preprocess code will be released soon.

• A part of financial news data can be found at ./raw_data/100E/financial news.xlsx and ./raw_data/300E/financial news.xlsx.

• The finance-oriented sentiment dictionary (found at ./raw_data/sentiment_dictionary) is used to extract sentiment from financial news.

inter-class relations

• The inter-class relations data can be found at ./raw_data/100E/inter-class and ./raw_data/300E/inter-class.

intra-class Relations

• The intra-class relations can be found at ./raw_data/100E/intra-class and ./raw_data/300E/intra-class.

Contact

[email protected]