• python 3.7.0
• torch==1.7.1
• scipy==1.7.0
• pandas==1.3.1
• numpy==1.20.3

MVSTT is implemented on those several public traffic datasets.

• PEMS03, PEMS04, PEMS07 and PEMS08 from STSGCN (AAAI-20).
• Due to the size limitation of the dataset, you can access our dataset in Google Cloud Drive.PEMS

PEMS03, PEMS04, PEMS07, PEMS08:

Run python train.py with parameters.

• --dataset: The datasets name.
• --train_rate: The ratio of training set.
• --seq_len: The length of input sequence.
• --pre_len: The length of output sequence.
• --batchsize: Number of training batches.
• --heads: The number of heads of multi-head attention.
• --dropout: Dropout.
• --lr: Learning rate.
• --in_dim: Dimensionality of input data.
• --embed_size: Embedded dimension.
• --epochs：The number of train epochs.

python train.py --dataset PEMS08

The pre-trained model is available in Google Cloud Drive.pre-trained model

PEMS03, PEMS04, PEMS07, PEMS08:

Run python test.py with parameters.

• --model: Trained model paths.

python test.py --dataset PEMS08 --model model/PEMS08/PEMS08.pkl