Soil Classification Based on Deep Learning Algorithm and Visible Near-Infrared Spectroscopy

Table 10

The classification results of soil samples based on convolutional neural network and support vector machine by randomly dividing the calibration set 10 times.

Number of times	1/3 proportion		1/4 proportion		1/5 proportion
	CNN (%)	SVM (%)	CNN (%)	SVM (%)	CNN (%)	SVM (%)
1	93.92	90.61	96.06	93.10	81.11	85.71
2	91.16	86.74	84.24	88.18	92.17	82.49
3	88.95	88.95	93.10	85.22	87.10	86.18
4	93.37	85.08	84.73	82.27	79.72	83.87
5	91.71	87.85	81.28	85.22	86.18	85.25
6	90.61	86.19	89.16	89.66	86.18	88.02
7	91.71	86.19	95.07	91.63	85.25	84.33
8	86.19	90.06	90.64	88.18	78.34	86.18
9	93.37	92.82	91.13	85.71	94.01	90.78
10	93.37	95.03	91.63	91.63	78.34	79.26
Average	91.44	88.95	89.70	88.08	84.84	85.21

CNN: convolutional neural network; SVM: support vector machine.