Nonlinear Multivariate Calibration of Shelf Life of Preserved Eggs (Pidan) by Near Infrared Spectroscopy: Stacked Least Squares Support Vector Machine with Ensemble Preprocessing

Table 2

Different preprocessing methods used for data fusion and the results of different models.

No.	Preprocessing^a	RMSEC (days)	RMSEMCCV (days)	RMSEP (days)
1	S (5, 2)^b	3.3	8.2	7.7
2	S (7, 2)	3.5	8.4	8.0
3	S (9, 2)	3.2	8.5	7.9
4	S (9, 3)	3.4	8.4	7.9
5	S (11, 3)	3.1	8.8	7.8
6	S (13, 3)	3.3	8.7	7.6
7	S (15, 3)	4.0	8.5	7.3
8	S (17, 4)	3.9	8.5	7.8
9	S (19, 4)	3.9	8.5	7.6
10	S (21, 5)	4.4	8.6	7.4
11	D1 (5, 2)	0.1	9.6	8.0
12	D1 (7, 2)	0.1	9.8	8.5
13	D1 (9, 2)	0.2	53.2	51.2
14	D1 (9, 3)	0.1	9.9	8.1
15	D1 (11, 3)	0.1	10.0	8.5
16	D1 (13, 3)	0.1	9.8	8.3
17	D1 (15, 3)	0.1	9.8	8.7
18	D1 (17, 4)	0.1	9.8	8.6
19	D1 (19, 4)	0.7	9.7	8.2
20	D1 (21, 5)	0.1	9.9	8.6
21	D2 (5, 2)	0.1	11.9	10.1
22	D2 (7, 2)	0.1	12.5	10.2
23	D2 (9, 2)	0.1	13.0	10.4
24	D2 (9, 3)	0.8	13.1	10.4
25	D2 (11, 3)	0.1	12.9	10.0
26	D2 (13, 3)	0.1	12.3	9.3
27	D2 (15, 3)	0.1	11.8	8.4
28	D2 (17, 4)	0.1	13.5	10.5
29	D2 (19, 4)	0.1	13.4	9.9
30	D2 (21, 5)	0.1	12.7	8.8
31	1 + SNV	3.2	9.1	8.3
32	2 + SNV	3.7	9.4	8.6
33	3 + SNV	3.7	9.6	8.6
34	4 + SNV	3.6	9.6	8.6
35	5 + SNV	3.5	9.7	8.5
36	6 + SNV	3.3	9.8	8.4
37	7 + SNV	3.4	9.7	8.1
38	8 + SNV	3.5	9.8	8.6
39	9 + SNV	3.5	9.8	8.4
40	10 + SNV	3.2	9.9	8.2
41	11 + SNV	1.1	9.7	8.5
42	12 + SNV	2.1	9.9	8.7
43	13 + SNV	1.4	9.8	8.9
44	14 + SNV	2.1	9.8	8.2
45	15 + SNV	1.7	9.9	8.7
46	16 + SNV	1.3	9.8	8.8
47	17 + SNV	1.3	9.9	9.2
48	18 + SNV	1.8	10.0	9.2
49	19 + SNV	1.9	9.9	8.8
50	20 + SNV	1.4	10.0	9.4
51	21 + SNV	0.1	12.7	10.4
52	22 + SNV	0.1	13.1	10.4
53	23 + SNV	0.1	12.7	10.4
54	24 + SNV	0.1	12.8	10.6
55	25 + SNV	0.1	11.9	9.8
56	26 + SNV	0.3	11.0	9.3
57	27 + SNV	0.7	10.5	8.6
58	28 + SNV	0.1	12.5	10.7
59	29 + SNV	0.1	12.1	10.0
60	30 + SNV	0.1	11.8	9.1
61	Raw data	3.2	8.1	7.6
62	SNV	3.0	8.9	8.1
63	Ensemble model	2.3	7.2	6.2

aS: smoothing, D1: taking first-order derivative, and D2: taking second-order derivative.
bThe numbers in the brackets indicate the number of data points and the order of fitting polynomial used in the algorithm by Savitzky and Golay.