Detecting extreme traffic events via a context augmented graph autoencoder
Yue Hu, Ao Qu and Dan Work, 2021
Context augmented Graph Autoencoder (Con-GAE) aims at detecting extreme events in traffic origin-destinatin (OD) datasets. Con-GAE leverages graph embedding and context embedding techniques to capture the spatial and temporal patterns in traffic dynamics, and adopts an autoencoder framework to detect anomalies via semi-supervised learning. Pre-processed Uber Movement data is also contained in this repository for tests.
The results are reported in "Detecting extreme traffic events via a context augmented graph autoencoder" by Y.Hu et. al. published at ACM Transactions on Intelligent Systems and Technology (TIST)
The current version of this code was written using python 3.8, Pytorch 1.11.0, Pytorch Geometric 2.0
Other required libraries include: numpy, scipy, pandas etc.
First chage directory to /src
To evaulate trained model, run:
python3 test.py --model 'ConGAE'
for main model ConGAE applied on NYC dataset.
To train model with configurations stated in paper, run:
python3 train.py --model 'ConGAE'
for main model ConGAE.
Full command for ConGAE training is:
python3 train.py --model 'ConGAE' --randomseed 1 '--train_epoch 150 --lr 1e-3 --dropout_p 0.2 --adj_drop 0.2 --verbal False --input_dim 4 --node_dim1 150 --node_dim2 50 --encode_dim 50 --hour_emb 100 --week_emb 100 --decoder 'concatDec'
ConGAE model has two graph convolutional layers. For deep model with more graph convolutional layers, use the deepConGAE variant:
python3 train.py --model 'deepConGAE' --hidden_list 300 150 150 --encode_dim 150 --decode_dim 150
where hidden_list contains the node embedding dimension of each layer of GCN