Automatic Traffic Anomaly Detection on the Road Network with Spatial-Temporal Graph Neural Network Representation Learning

Table 5

Results of anomaly detection compared with baseline models.

Method	Accuracy	Detection rate	False alarm rate
LOF	0.80	0.89	0.09
IFOREST	0.83	0.80	0.08
OCSVM	0.82	0.84	0.08
HBOS	0.87	0.31	0.03
COPOD	0.82	0.82	0.08
LSTM-VAE	0.97	0.97	0.03
MAD-GAN	0.91	0.35	0.16
GDN (the selected model)	0.97	0.97	0.02