Multivariate and Propagation Graph Attention Network for Spatial-Temporal Prediction with Outdoor Cellular Traffic
This is the original mobile cellular traffic dataset of MP-GAT in the following paper: Multivariate and Propagation Graph Attention Network for Spatial-Temporal Prediction with Outdoor Cellular Traffic (CIKM 2021). NOTE: All the cellular record is anonymized by hashing, then aggregate the quantity as our dataset.
1.Road Intersection distance and the Corresponding Adjacent Matrix Google Drive
The cellular traffic data files for Hsinchu City, Taiwan, i.e., cellular_traffic_202001_202006.csv
, are available at Google Drive.
Here is an example:
Intersection_0 | Intersection_1 | Intersection_2 | Intersection_n | |
---|---|---|---|---|
2020/01/01 00:00:00 | 795.0 | 124.0 | 175.0 | ... |
2020/01/01 00:05:00 | 750.0 | 141.0 | 159.0 | ... |
2020/01/01 00:10:00 | 538.0 | 73.0 | 178.0 | ... |
... | ... | ... | ... | ... |
To generate train/test/val dataset as {train,val,test}.npz
, please look a nice reference script from DCRNN.
The rule of Mean Average 5/20 and Daily cellular traffic, please refer to our paper in section 3: PRELIMINARIES.