wide-attention and deep model

• Python 2.7
• Tensorflow-gpu 1.5.0
• Keras 2.1.3
• scikit-learn 0.19

Run command below to train the model:

• Train the baseline single DL model based on CPeMS dataset.

python train_t.py --model model_name

You can choose "lstm", "gru" or "saes" as arguments. The .h5 weight file was saved at model folder.

• Train the composite DL model based on CPeMS dataset.

python train_wd.py --model model_name

You can choose "w_attention_d" (WADM) or "wd_crossLayer_attention" (DCN) as arguments. The .h5 weight file is saved at model folder.

• Training model based on FBBC dataset.

python train_bike.py --model model_name

You can choose "lstm", "gru" as arguments for training single DL model or choose "w_attention_d" (WADM) for training composite DL model.

Data are obtained from the Caltrans Performance Measurement System (CPeMS) and Fremont Bridge Bicycle Counter (FBBC).

device: GTX 1050
dataset: CPeMS and FBBC
optimizer: RMSprop

@ARTICLE{9120076,
  author={J. {Zhou} and H. {Dai} and H. {Wang} and T. {Wang}},
  journal={IEEE Transactions on Industrial Informatics}, 
  title={Wide-Attention and Deep-Composite Model for Traffic Flow Prediction in Transportation Cyber-Physical Systems}, 
  year={2020},
  volume={},
  number={},
  pages={1-1},}