Quantitative analysis of Mn on Mars from SuperCam⁃LIBS spectral datasets

Fig. 8. Comparison of the predicted Mn content of each model with the true value. The 1:1 reference line means that the predicted value is the same as the true value, and the closer the prediction result is to the reference line,the more accurate the prediction is for the model

Download full size

View in Article

Table 1. Compositional information of SuperCam samples

View table

View in Article

Table 1. Compositional information of SuperCam samples

No.	Rock type	Number of samples	Composition of MnO/%
1	Mn ore	7	34.62~76.19
2	Manganese nodule	5	16.52~33.09
3	Basalt	56	0.01~4.92
4	Phyllosilicate	11	0.08~0.59
5	Olivine	4	0.09~0.42
6	Augite	2	0.22~0.24
7	Plagioclase	25	0.05~0.18
8	Fe-oxide	6	0.03~1.95
9	Others	171	0.01~1

Table 2. Fitting parameters and R² for three univariate models
View table
View in Article
Table 2. Fitting parameters and R² for three univariate models
Fitting type Fitting parameters R²
$y = a x$ $a$ =0.00491 0.82
$y = a x + b x^{2}$ $a$ =0.01129, $b$ =-0.000115 0.89
$y = a x^{b}$ $a$ =0.03309, $b$ =0.51984 0.91

Table 3. Model parameters
View table
View in Article
Table 3. Model parameters
Model name Parameters
GBDT N_estimators=100
RF N_estimators=100
LGBM N_estimators=100
SVR C=1000
Gamma=0.03
LASSO Alpha=0.000843
ElasticNet Alpha=0.000968
L1=1

Table 4. Validation of the models at different elemental content ranges

View table

View in Article

Table 4. Validation of the models at different elemental content ranges

Training set minimum (MnO/%)	Training set maximum (MnO/%)	Method	E_rms/%	R²
0	70	Stacking	2.72	0.94
		LASSO	5.34	0.86
		ElasticNet	3.87	0.92
0	1	Stacking	0.025	0.96
		LASSO	0.063	0.94
		ElasticNet	0.103	0.91
10	70	Stacking	5.96	0.9
		LASSO	10.78	0.82
		ElasticNet	7.74	0.89

Table 5. Comparison of true and predicted values of some samples under each model

View table

View in Article

Table 5. Comparison of true and predicted values of some samples under each model

Sample Name	Rock Type	Method		MnO Composition/%
JMN1	Mn nodule	Stacking	True	33.09
		Stacking	Predicted	34.7
		LASSO	True	33.09
		LASSO	Predicted	61
		ElasticNet	True	33.09
		ElasticNet	Predicted	35.1
PMIFS0301	Olivine	Stacking	True	0.42
		Stacking	Predicted	0.41
		LASSO	True	0.42
		LASSO	Predicted	0.37
		ElasticNet	True	0.42
		ElasticNet	Predicted	0.28
JSC1380	Basalt	Stacking	True	0.1
		Stacking	Predicted	0.105
		LASSO	True	0.1
		LASSO	Predicted	0.07
		ElasticNet	True	0.1
		ElasticNet	Predicted	0.15

Tools

Get Citation

Copy Citation Text

Mingyu SU, Yanqing XIN, Changqing LIU, Zongcheng LING. Quantitative analysis of Mn on Mars from SuperCam⁃LIBS spectral datasets[J]. Chinese Journal of Quantum Electronics, 2024, 41(3): 473

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites