3D-GCN

This is the data and code for submission#683.

Dataset

Our dataset is from the NYC Bike. We obtained road network and 9 categories of PoIs from OpenStreatMap. We devided the city into 82 irregular regions, based on road network.

The dataset contains 2208 time intervals, spanning from Jul. 1st to Sept. 30th, 2017 (92 days, 24 intervals per day). You can download the dataset via following links: Jul, 2017 Aug, 2017 Sept, 2017

Requirements

• python 2.7
• PyTorch 0.2.0
• NumPy
• JSON

Project Structure

• /data
  • flow_bike_nyc_irregular.json contains the inflows, outflows obtained from the dataset
  • irregular_feature.npy is the poi features of all regions
  • irregular_idx.npy is the index of labeled regions
  • irregular_label.npy is the labels of all regions
  • irregular_weight.npy is the normalization term
  • irregular_path.npy contains the historical OD flows for the irregular dataset
• /code
  • main.py run this file to get the results of 3D-GCN in our paper
  • gcn.py is the implementation of our neural network
  • dataset.py loads the dataset

Usage

python main.py