Discrimination of Boletaceae mushrooms based on data fusion of FT-IR and ICP–AES combined with SVM

Figures & data

Table 1. Information of the mushroom samples.

Code	No.	Latin name
A	1–30	Boletus speciosus
B	31–78	Leccinum rugosiceps
C	79–123	Boletus umbriniporus
D	124–148	Boletus tomentipes
E	149–230	Boletus edulis

Figure 1. Raw FT-IR spectrum (a) and spectrum preprocessed by 2D + SNV + SG (15:1) (b) of different species wild-grown Boletaceae mushrooms.

Table 2. The results of mineral concentrations determined in five species mushrooms (mean ± SD, mg kg⁻¹).

	A	B	C	D	E
Ba	67.92 ± 36.95	27.80 ± 25.76	25.06 ± 29.10	32.18 ± 12.09	79.28 ± 58.99
Ca	149.04 ± 61.68	118.55 ± 65.10	122.97 ± 118.81	100.38 ± 41.39	317.19 ± 183.52
Cd	6.87 ± 4.16	4.65 ± 5.78	5.46 ± 6.03	9.10 ± 5.78	21.07 ± 64.58
Co	11.97 ± 16.95	3.64 ± 3.10	3.94 ± 6.00	4.08 ± 2.41	7.34 ± 5.83
Cr	74.34 ± 61.70	47.10 ± 33.98	21.35 ± 22.34	31.11 ± 33.25	87.53 ± 137.94
Cu	23.16 ± 16.61	22.22 ± 11.54	40.04 ± 44.47	42.88 ± 48.24	33.47 ± 29.01
Fe	976.42 ± 630.18	459.10 ± 437.76	517.12 ± 334.02	568.80 ± 275.17	1759.01 ± 1514.30
K	5938.66 ± 2070.22	5944.55 ± 2915.92	8662.21 ± 2873.45	11015.43 ± 1083.45	9269.86 ± 3375.22
Li	5.19 ± 4.60	1.35 ± 1.51	1.16 ± 1.86	0.40 ± 0.70	6.78 ± 11.33
Mg	554.53 ± 147.57	275.91 ± 142.81	327.29 ± 175.21	369.38 ± 126.53	400.20 ± 233.42
Mn	19.89 ± 11.31	28.38 ± 18.25	21.00 ± 14.71	30.38 ± 10.58	43.63 ± 24.71
Na	200.78 ± 220.02	116.30 ± 10.63	316.09 ± 252.08	241.84 ± 105.21	181.90 ± 230.32
Ni	24.51 ± 19.95	10.63 ± 17.76	48.02 ± 81.19	24.82 ± 35.10	42.83 ± 43.71
P	2721.39 ± 1044.72	2549.64 ± 957.74	2755.57 ± 1355.51	2607.33 ± 564.19	3217.08 ± 1206.51
Rb	319.13 ± 212.20	297.32 ± 281.32	339.74 ± 214.90	217.75 ± 215.44	1115.93 ± 991.18
Sr	9.70 ± 3.67	4.91 ± 2.72	8.19 ± 7.69	3.69 ± 2.91	31.81 ± 49.61
V	22.21 ± 13.72	7.61 ± 8.24	9.57 ± 8.21	6.20 ± 6.89	29.41 ± 28.54
Zn	79.46 ± 50.96	49.13 ± 31.05	59.86 ± 35.46	89.59 ± 16.13	66.79 ± 39.41

Figure 2. Latent variables number and R²Y, Q² values.

Table 3. The optimization results for parameters c and g by grid-search method.

Data matrices	c	g	Cross-validation accuracy (%)
FT-IR spectrum	1351.18	0.0039063	96.75
Element concentrations	16	0.32988	88.96
Low-level data fusion	445.7219	0.00014	95.45
Mid-level data fusion	0.57435	0.0625	100.00

Figure 3. The optimization results for parameters (c and g) and cross-calidation accuracy by grid-search method with mid-level data fusion strategy. A: FT-IR spectra; B: element concentrations.

Table 4. The results of SVM models on different data matrices.

Data matrices	Accuracy of cross-validation (%)	Accuracy of test set (%)
FT-IR spectrum	96.75	88.16
Element concentrations	88.96	73.68
Low-level data fusion	95.45	94.74
Mid-level data fusion	100.00	98.68

Figure 4. The plot of actual and predicted category of test samples by support vector machine.

A: FT-IR spectra; B: element concentrations; C: low-level data fusion; D: mid-level data fusion.