Table 1 Results for feature selection.

Statistics in SVM model	Overall classification accuracy (%)
Statistics in SVM model	SHG SVM model	1450 cm⁻¹ SRS SVM model	1668 cm⁻¹ SRS SVM model
All parameters	92.60	49.30	85.40
No mean	88.08	44.59	87.36*
No variance	92.61*	34.50	55.95
No skewness	93.00*	46.88	83.23
No kurtosis	92.94*	47.10	85.44*
No contrast	92.14	47.35	84.90
No correlation	89.18	46.64	82.78
No energy	93.44*	46.90	85.11
No homogeneity	79.92	48.37	84.74
No entropy	91.92	44.72	85.15

Mean classification accuracy of the central diagonal of the confusion matrix representing subimage accuracy as a function of which parameters are trained on. Italicised quantities with an asterisk are classification accuracies that are higher than when using all parameters (shown in bold), indicating that omitting that statistic increased the classification accuracy of the model.

Quick links

Search