Detecting Suspects by Large-Scale Trajectory Patterns in the City

Table 3

Experimental results with different classification methods.

Model	Methods	Accuracy	Precision	Recall	F3 measure
RF	SPM	0.818	0.009	0.602	0.079
	LCM	0.867	0.028	0.786	0.212
	TPM	0.912	0.021	0.694	0.165
	LCM + TPM	0.915	0.042	0.96	0.301
NB	SPM	0.852	0.007	0.714	0.064
	LCM	0.861	0.018	0.726	0.147
	TPM	0.923	0.015	0.857	0.114
	LCM + TPM	0.935	0.014	0.818	0.121
GBDT	SPM	0.807	0.012	0.654	0.103
	LCM	0.844	0.029	0.785	0.218
	TPM	0.956	0.022	0.703	0.172
	LCM + TPM	0.962	0.051	0.972	0.346
LR	SPM	0.844	0.009	0.651	0.08
	LCM	0.851	0.013	0.673	0.111
	TPM	0.892	0.018	0.664	0.145
	LCM + TPM	0.906	0.019	0.886	0.159
KNN	SPM	0.837	0.014	0.71	0.119
	LCM	0.852	0.034	0.71	0.238
	TPM	0.869	0.015	0.69	0.125
	LCM + TPM	0.950	0.035	0.86	0.256
OSCVM	SPM	0.907	0.018	0.912	0.153
	LCM	0.912	0.019	0.995	0.162
	TPM	0.909	0.018	0.932	0.154
	LCM + TPM	0.921	0.017	0.930	0.146
Deep learning	MST-CNN	0.981	0.191	0.650	0.524