Evaluation Models for Soil Nutrient Based on Support Vector Machine and Artificial Neural Networks

Table 2

Computing results of the two models in different normalization conditions.

Model (trained for 1000 times)	Proportion of training set	Computing results	No normalization	−1, 1 Normalization	0, 1 Normalization
Model 1	35%	Average prediction accuracy	77.87%	80.91%	82.34%
		Standard deviation	0.1396	0.1911	0.1727
Model 2	20%	Average prediction accuracy	83.00%	71.53%	72.75%
		Standard deviation	0.1463	0.2548	0.2464