Forest Ecosystems

Table 6 Confusion matrix from cross-validation of the four artificial neural network model results

From: Evaluating soil nutrients of Dacrydium pectinatum in China using machine learning techniques

	Grade	Measured grade (%)
	Grade	I	II	III	IV	V	VI
Predicted grade	I	15.4	0.5	1.0	1.7	1.0	0.8
	II	1.2	15.5	2.0	1.2	0.5	0.5
	III	0.8	0.7	15.1	2.0	0.4	1.1
	IV	0.9	1.1	1.7	15.7	0.8	0.6
	V	1.2	1.3	0.9	0.3	15.2	0.5
	VI	0.3	0.4	0.6	0.5	0.8	15.6
Accuracy		92.5
Predicted grade	I	15.7	1.3	1.5	0.6	0.8	1.6
	II	0.4	15.5	1.2	0.7	0.9	0.3
	III	0.3	2.1	15.2	0.9	1.3	0.4
	IV	0.3	1.5	1.8	15.3	0.4	0.8
	V	0.7	0.8	1.2	0.8	15.4	0.5
	VI	0.5	0.7	0.7	0.6	0.4	14.9
Accuracy		92.0
Predicted grade	I	15.7	0.8	1.1	0.7	0.9	0.5
	II	1.7	14.8	0.7	0.5	0.6	1.4
	III	0.6	0.7	14.4	0.6	1.8	1.1
	IV	0.3	0.6	0.3	15.5	1.2	1.6
	V	0.8	1.7	0.2	0.9	14.7	1.4
	VI	0.9	1.5	0.5	0.4	0.6	14.4
Accuracy		89.5
Predicted grade	I	14.7	0.6	1.2	0.9	1.1	0.7
	II	1.4	14.4	0.5	0.7	0.8	1.6
	III	0.7	0.9	15.0	0.7	1.3	1.0
	IV	0.8	0.8	0.6	14.9	1.6	1.4
	V	0.5	1.3	0.3	0.6	15.2	1.3
	VI	0.7	1.7	0.4	0.9	0.7	14.3
Accuracy		88.5

Note: I, II, III, IV, V, and VI indicate that the soil nutrient quality grade is “extremely high”, “medium high”, “low”, “poor”, and “extremely poor”, respectively. Four models are back propagation neural network (BPNN), field probing neural networks (FPNN), muhilayer perceptron neural networks (MLPNN), and general regression neural network (GRNN)

Back to article page