Traffic Flow Prediction and program codes

Citation Author(s):
Hong-Ning
Dai
Macau University of Science and Technology
Junhao
Zhou
Macau University of Science and Technology
Submitted by:
Hong-Ning Dai
Last updated:
Mon, 06/15/2020 - 05:15
DOI:
10.21227/ac0c-8043
Data Format:
Links:
License:
0
0 ratings - Please login to submit your rating.

Abstract 

This dataset mainly consists 1) source codes of wide-attention and deep model (WADC); 2) datasets to evaluate the performance of the proposed model. Datasets are obtained from the Caltrans Performance Measurement System (CPeMS) http://pems.doc.ca.gov; and Fremont Bridge Bicycle Counter (FBBC), https://data.seattle.gov.

 

For more details, please refer to our paper entitled "Wide-Attention and Deep-Composite Model for Traffic Flow Prediction in Transportation Cyber-Physical Systems", accepted to appear in IEEE Transactions on Industrial Informatics and github https://github.com/zhoujunhao/wadc.

Instructions: 

WADC

wide-attention and deep model

Installation

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

Train the model

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.

Experiment

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