Github: https://github.com/SuminHan/Traffic-UAGCRNTF

We share our code for deep-learning research from Sumin Han, Youngjun Park, Minji Lee, Jisun An, and Dongman Lee, “Enhancing Spatio-temporal Traffic Prediction through Urban Human Activity Analysis,” has been accepted to 32nd ACM International Conference on Information and Knowledge Management Paper Code

This code contains not only our proposed UAGCRN and UAGCTransformer, but also DCRNN and GMAN which are baselines.

Model Architecture

Code

This installation command worked with our code execution:

conda deactivate
conda env remove -n CIKM
conda create -n CIKM -y
conda activate CIKM 
conda install python==3.10 -y
pip install tensorflow-gpu==2.10.0
pip install tqdm
pip install tables
pip install scipy==1.10.1
pip install pandas==1.5.3
pip install numpy==1.23.5

python train.py --model_name=MyUAGCRN --dataset=metr-la --Q=12 --activity_embedding --sensor_embedding --graph_type=cooccur_dist
python train.py --model_name=MyUAGCRN --dataset=pems-bay --Q=12 --activity_embedding --sensor_embedding --graph_type=cooccur_dist
python train.py --model_name=MyUAGCRN --dataset=pemsd7 --Q=9 --activity_embedding --sensor_embedding --graph_type=cooccur_dist
python train.py --model_name=MyUAGCTransformer --dataset=metr-la --Q=12 --activity_embedding --sensor_embedding --graph_type=cooccur_dist
python train.py --model_name=MyUAGCTransformer --dataset=pems-bay --Q=12 --activity_embedding --sensor_embedding --graph_type=cooccur_dist
python train.py --model_name=MyUAGCTransformer --dataset=pemsd7 --Q=9 --activity_embedding --sensor_embedding --graph_type=cooccur_dist

Dataset

Traffic Dataset

california-vds.json: https://pems.dot.ca.gov/
metr-la, pems-bay: DCRNN, https://github.com/liyaguang/DCRNN
pemsd7: STGCN, https://github.com/VeritasYin/STGCN_IJCAI-18

Traffic Dataset Description

DATASET	N	Speed (miles/hour)	Datasize	PLACE	DURATION	INTERVAL
METR_LA	207	54 ± 20	34,249	Los Angeles, USA	Mar. 1, 2012 – Jun. 27, 2012	5min
PEMS_BAY	325	62 ± 10	52,093	San Francisco Bay Area, USA	Jan. 1, 2017 – Jun. 30, 2017	5min
PEMSD7	228	59 ± 13	12,652	Los Angeles, USA	May. 1, 2012 – Jun.30, 2012	5min

National Household Survey

Citation

To recognize the valuable role of National Household Travel Survey (NHTS) data in the transportation research process and to facilitate repeatability of the research, users of NHTS data are asked to formally acknowledge the data source. Where possible, this acknowledgement should take place in the form of a formal citation, such as when writing a research report, planning document, on-line article, and other publications. The citation can be formatted as follows: U.S. Department of Transportation, Federal Highway Administration, 2017 National Household Travel Survey. URL: http://nhts.ornl.gov.