beginner-sketch / MoSSL

[IJCAI 2024] This it the official github for IJCAI24 paper "Multi-Modality Spatio-Temporal Forecasting via Self-Supervised Learning"

[IJCAI 2024] Multi-Modality Spatio-Temporal Forecasting via Self-Supervised Learning

[IJCAI24] Jiewen Deng, Renhe Jiang, Jiaqi Zhang, Xuan Song, "Multi-Modality Spatio-Temporal Forecasting via Self-Supervised Learning", IJCAI, 2024.

Our research has been accepted for presentation at the main track of IJCAI 2024.

This implementation showcases our MoSSL model.

Preprint Link

Citation

Citation details will be updated once the official proceedings for IJCAI 2024 are available online.

@article{deng2024multimodality,
  title={Multi-Modality Spatio-Temporal Forecasting via Self-Supervised Learning},
  author={Jiewen Deng and Renhe Jiang and Jiaqi Zhang and Xuan Song},
  journal={arXiv preprint arXiv:2405.03255},
  year={2024}
}

Multi-Modality Spatio-Temporal Dataset

NYC Traffic Demand dataset^[1] is collected from the New York City, which consists of 98 nodes and four transportation modalities: Bike Inflow, Bike Outflow, Taxi Inflow, and Taxi Outflow. The timespan is from April 1st, 2016 to June 30th, 2016, and the time interval is set to half an hour.

BJ Air Quality dataset^[2] is collected from the Beijing Municipal Environmental Monitoring Center, which contains 10 nodes and three pollutant modalities: $PM_{2.5}$, $PM_{10}$, and $SO_2$. The timespan is from March 1st, 2013 to February 28th, 2017, and the time interval is set to one hour.

1. ^ https://ride.citibikenyc.com/system-data; https://www1.nyc.gov/site/tlc/about/tlc-trip-record-data.page

2. ^ https://archive.ics.uci.edu/ml/datasets/Beijing+Multi-Site+Air-Quality+Data

Installation Dependencies

Python 3 (>= 3.6; Anaconda Distribution)

PyTorch (>= 1.6.0) 

Numpy >= 1.17.4

Pandas >= 1.0.3

Model Training and Testing

python traintest_MoSSL.py cuda_name