Provenance classification of nephrite jades using multivariate LIBS: a comparative study

Provenance classification of nephrite jades is important since the unit price of jade changes drastically with its geological origin. In the present work, a detailed comparison between commonly applied multivariate methods is conducted to classify nephrite samples from five different locations via their laser-induced breakdown spectroscopy (LIBS) spectra. Five multivariate methods including principal component analysis (PCA), one-step and pairwise partial least squares discriminant analysis (PLS-DA), linear discriminant analysis (LDA), and support vector machine (SVM) were applied to provide the classification information on the samples. PCA was used for rough classification while the classification performance of the other four methods was discussed in detail. The results show a training accuracy of 89.0%, 99.2%, 99.8% and 100%, and a testing accuracy of 76.8%, 97.8%, 92.8% and 99.3% for the PLS-DA (one-step), PLS-DA (pairwise), LDA and SVM algorithms respectively. The superior model nature and the selection of suitable characteristic lines for weight differences led to the high performance of SVM, showing an excellent applicability for the provenance classification of nephrite jades using LIBS spectra.