Stock Prediction using Graph Neural Networks (GNN): Benchmarking relational graphs of SP500 companies
Grade: 10/10
In this thesis, a benchmark is developed for the evaluation of relational graphs of SP500 companies for the stock market prediction problem. Multiple methods of constructing relational graphs are evaluated with industry sources, news, structure, institutional investors, correlations of indicators, etc. To develop the benchmark, a heterogeneous temporal Graph Neural Network has been reconstructed, which is able to beat the market consistently over time. Furthermore, a method of evaluating the graph by taking into account the correlation graph of the following period has been proposed. In this way, the similarity between graphs is measured and can facilitate the investigation of relational graph construction methods in future research.
First of all, the feature data and network generation must be generated. Since these files can be very heavy, they could not be included in the repository. In the crawlers folder they can be generated.
After checking the requirements.txt file, the GNN can be executed through the scheduler.py file.
-
Author: Miquel Muñoz
miquelmunozz@gmail.com -
Director: Sergi Abadal
sergiabadal@gmail.com -
Codirector: Axel Wassington
axelwass@gmail.com
This thesis has reconstructed an existing architecture of a heterogeneous graph neural network. The LICENSE of the project cited is GPL3, which means no matter how you modify or use code, you need to open source it.
@inproceedings{Xiang2022Temporal,
author = {Xiang, Sheng and Cheng, Dawei and Shang, Chencheng and Zhang, Ying and Liang, Yuqi},
title = {Temporal and Heterogeneous Graph Neural Network for Financial Time Series Prediction},
year = {2022},
isbn = {9781450392365},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3511808.3557089},
doi = {10.1145/3511808.3557089},
booktitle = {Proceedings of the 31st ACM International Conference on Information & Knowledge Management},
pages = {3584–3593},
numpages = {10},
location = {Atlanta, GA, USA},
series = {CIKM '22}
}