Table 6.

Random forest algorithm results for single layer stacked image.

Random forest for single image
Training 70							Testing 30
Maximum depth		Fields_ training	Forest_ training	Shrubs_ training	Urban_ training	Overall. accuracy	Fields_ testing	Forest_ testing	Shrubs_ testing	Urban_ testing	Overall. accuracy
5	Overall accuracy = 88.43%, kappa coefficient = 0.793						Overall accuracy = 91.00%, kappa coefficient = 0.846
	Unclassified	0	0	0	0		0	0	0	0
	Fields	14,892	1317	488	403	84.22%	6473	515	220	168	88.14%
	Forest	637	31,320	1035	52	91.89%	194	13,469	475	7	94.7%
	Shrubs	1328	1449	2727	0	64.06%	283	239	1535	0	68.65%
	Urban	826	0	7	8708	95.03%	394	0	6	3790	95.59%
10	Overall accuracy = 92.83%, kappa coefficient = 0.881						Overall accuracy = 92.58%, kappa coefficient = 0.877
	Unclassified	0	0	0	0		0	0	0	0
	Fields	15,922	539	434	205	90.20%	6654	320	239	163	91.35%
	Forest	482	31,896	604	62	95.14%	171	13,590	373	11	95.51%
	Shrubs	752	1089	3663	0	77.80%	174	306	1577	0	71.84%
	Urban	496	2	7	9036	97.13%	285	12	6	3887	95.71%
20	Overall accuracy = 99.12%, kappa coefficient = 0.986						Overall accuracy = 92.90%, kappa coefficient = 0.882
	Unclassified	0	0	0	0		0	0	0	0
	Fields	17,027	20	50	3	98.09%	6706	273	226	171	92.55%
	Forest	67	32,934	43	0	99.57%	157	13,664	310	14	95.13%
	Shrubs	122	123	5259	0	98.26%	172	411	1474	0	73.00%
	Urban	143	0	0	9398	99.97%	211	15	9	3955	95.53%