Table 4.

Comparative retrieval performance with and without predicting the class label (CL).

Retrieval and data sets		F1	APa	ARb	ACCc
Without class prediction
	MCd	0.844	0.861	0.828	0.847
	SZe	0.891	0.892	0.89	0.89
	MC + SZ	0.88	0.882	0.878	0.879
	MC train and SZ test	0.534	0.538	0.53	0.533
	SZ train and MC test	0.729	0.737	0.72	0.739
With class prediction
	MC	0.929	0.937	0.921	0.928
	SZ	0.908	0.909	0.908	0.908
	MC + SZ	0.9	0.902	0.898	0.899
	MC train and SZ test	0.795	0.798	0.793	0.792
	SZ train and MC test	0.811	0.808	0.813	0.797

^aAP: average precision.

^bAR: average recall.

^cACC: accuracy.

^dMC: Montgomery County.

^eSZ: Shenzhen.