Skip to content

Latest commit

 

History

History
49 lines (29 loc) · 1.65 KB

README.md

File metadata and controls

49 lines (29 loc) · 1.65 KB

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

| Introduction

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

| Citation

[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}
}

| Installation

pip install -r requirements.txt

| Datasets

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.

| Training and Testing

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

| Recreate plots in the paper

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