Wavelength Selection Algorithm Based on Minimum Correlation Coefficient for Multivariate Calibration

Jie-hong CHENG; Zheng-guang CHEN; Shu-juan YI

doi:10.3964/j.issn.1000-0593(2022)03-0719-07

Journals >Spectroscopy and Spectral Analysis >Volume 42 >Issue 3 >Page 719 > Article

Spectroscopy and Spectral Analysis
Vol. 42, Issue 3, 719 (2022)

Wavelength Selection Algorithm Based on Minimum Correlation Coefficient for Multivariate Calibration

Jie-hong CHENG^1、*, Zheng-guang CHEN^{1、1; 2; *;}, and Shu-juan YI^2、2;

Author Affiliations

¹1. College of Information and Electrical Engineering, Heilongjiang Bayi Agricultural University, Daqing 163319, China

²2. Heilongjiang Engineering Technology Research Center for Rice Ecological Seedlings Device and Whole Process Mechanization, Daqing 163319, China

show less

DOI: 10.3964/j.issn.1000-0593(2022)03-0719-07 Cite this Article

Jie-hong CHENG, Zheng-guang CHEN, Shu-juan YI. Wavelength Selection Algorithm Based on Minimum Correlation Coefficient for Multivariate Calibration[J]. Spectroscopy and Spectral Analysis, 2022, 42(3): 719 Copy Citation Text

show less

Fig. 1. Average value and standard deviation of correlation coefficient

Download full size | View in the Article

Fig. 2. Graphic analysis of threshold in grid T

Download full size | View in the Article

Fig. 3. Original (a) and Preprocessed (b) NIR spectra of diesel fuels sample

Download full size | View in the Article

Fig. 4. Wavelength selection results by MCC

Download full size | View in the Article

Fig. 5. Mean and standard deviation of correlation coefficient between selected wavelengths

Download full size | View in the Article

Fig. 6. Changes in RMSEV value of MCC-MLR model under different threshold coefficients

Download full size | View in the Article

相关系数绝对值	相关强度
0.8~1.0	极强相关
0.6~0.8	强相关
0.4~0.6	中等程度相关
0.2~0.4	弱相关
0.0~0.2	极弱相关或无相关

Table 1. The strength of correlation corresponding to the correlation coefficient between variables

View in the Article

性质	样本集	样本数	最小值	最大值	平均值	标准差	变异系数/%
Dataset1(BP50)	Whole set	246	197	293	258.38	17.06	6.6
	Calibration set	148	197	293	256.32	19.19	7.5
	Validation set	49	201	280	260.86	12.83	4.9
	Prediction set	49	198	283	262.12	12.62	4.8
Dataset2(SOM)	Whole set	108	42.91	95.85	85.43	10.82	12.7
	Calibration set	65	42.91	95.85	82.52	12.82	15.5
	Validation set	22	76.74	93.24	88.30	4.79	5.4
	Prediction set	21	86.65	93.46	91.40	1.63	1.8

Table 2. Results of sample chemical property

View in the Article

No.	建模方法	波长数	Validation set		Prediction set
No.	建模方法	波长数	R²	RMSE	R²	RMSE
1	FULL-PLSR	401	0.865 0	4.677 8	0.905 2	3.946 8
2	SPA-MLR	25	0.953 6	2.834 5	0.953 9	2.844 9
3	CARS-PLSR	18	0.928 9	4.760 1	0.927 9	3.544 5
4	IRIV-PLSR	60	0.966 4	3.274 7	0.956 8	2.639 6
5	RF-PLSR	34	0.948 0	4.071 9	0.943 4	3.009 9
6	MCC-MLR	20	0.953 9	2.760 1	0.956 0	2.779 2

Table 3. Model results based on different wavelength selection methods for diesel fuels datasets

View in the Article

No.	建模方法	波长数	Validation set		Prediction set
No.	建模方法	波长数	R²	RMSE	R²	RMSE
1	FULL-PLSR	850	0.585 2	3.229 9	0.656 2	2.917 6
2	SPA-MLR	23	0.953 9	0.908 9	0.810 1	1.870 9
3	CARS-PLSR	9	0.975 5	1.803 9	0.908 8	1.368 2
4	IRIV-PLSR	33	0.912 4	1.182 5	0.910 6	1.027 9
5	RF-PLSR	25	0.898 8	1.282 7	0.807 4	2.108 4
6	MCC-MLR	12	0.936 1	1.144 7	0.926 5	1.032 3

Table 4. Model results based on different wavelength selection methods for soil datasets

Download Citation

Save the article for my favorites

Paper Information