DYCOR: Capturing Hidden Stock Relationships for Stock Trend Prediction

📄 Paper PDF | 🏛️ ACM Digital Library

Official code implementation and supplementary material of CIKM 2025 paper "DYCOR: Capturing Hidden Stock Relationships for Stock Trend Prediction". This work proposes a novel stock trend prediction method that integrates dynamic stock clustering and correlation-aware training to capture evolving and latent relationships between stocks, addressing the limitations of existing methods that rely on predefined static relationships and fail to adapt to changing market dynamics.

Requirements

• Python >= 3.8
• See requirements.txt for package dependencies

Installation

git clone https://github.com/FinancialDeepLearning/DYCOR.git
pip install -r requirements.txt

Datasets

We evaluate our model on three benchmark datasets:

• NASDAQ: 1,026 stocks (Jan 2013 - Dec 2017)
• NYSE: 1,737 stocks (Jan 2013 - Dec 2017)
• S&P 500: 646 stocks (Jan 2003 - Dec 2023)

The NASDAQ and NYSE datasets are obtained from Temporal Relational Stock Ranking.

Usage

Replace NASDAQ with NYSE or SP500 for different datasets

cd src
python main.py --market NASDAQ --gpu 0

Citation

If you find this work useful for your research, please cite:

@inproceedings{choi2025dycor,
  title={DYCOR: Capturing Hidden Stock Relationships for Stock Trend Prediction},
  author={Choi, Kangmin and Shin, Geon and Yang, Jungwoo and Kim, Hyunjoon},
  booktitle={Proceedings of the 34th ACM International Conference on Information and Knowledge Management (CIKM '25)},
  year={2025},
  month=nov,
  address={Seoul, Republic of Korea},
  publisher={Association for Computing Machinery},
  doi={10.1145/3746252.3761413},
  isbn={979-8-4007-2040-6}
}