Copula Conformal Prediction algorithm for multivariate, multi-step Time Series (CopulaCPTS) is a conformal prediction algorithm with full-horizon validity guarantee.

[2212.03281] Copula Conformal Prediction for Multi-step Time Series Forecasting

@inproceedings{sun2023copula,
  title={Copula Conformal prediction for multi-step time series prediction},
  author={Sun, Sophia Huiwen and Yu, Rose},
  booktitle={The Twelfth International Conference on Learning Representations},
  year={2023}
}

pip install -r requirements.txt

Please see below for links and refer to Section 5.1 and Appendix C.1 in the paper for processing details.

Particles | Drone| Epidemiology | Argoverse 1

The processed files for Particles, Drone, and Epidemiology datasets are located in the ./data directory. If you want to reporduce the visualizations, you might need to refer to the original sources for metadata.

To illustrate the usage of our code, we have included pre-generated NRI Particles data in this repository. To replicate the experiment, simply run:

./run_experiment.sh

Please see Visualization.ipynb for example code for creating Figure 3 in the paper.