Heterogeneous and Dynamic Aware Spatio-Temporal Learning for Route-based Travel Time Estimation

This a framework called HDTTE to incorporate the heterogeneous and dynamic information of spatio-temporal traffic for effective travel time estimation.

Environment

• python 3.9.7
• torch 1.2.0
• numpy 1.17.2

Dataset

Step 1： Download the processed Wuhan dataset from Baidu Yun

If needed, the origin dataset of PEMSD4 and PEMSD8 are available from ASTGCN.

Step 2: Put them into data directories.

Train command

# Train with Wuhan dataset
python run.py --data=WUHAN --is_train

# Train with Beijing dataset
python run.py --data=BJ --is_train