The Development of Python Based Routine for Material Classification Using Laser Induced Breakdown Spectroscopy Data

Abstract

Laser Induced Breakdown Spectroscopy (LIBS) is a highly capable tool for diverse applications. One of its significant uses is the classification of materials in various categories ranging from pure metals to organic samples. It can be accomplished with high success by using machine learning classification algorithms. LIBS coupled with support vector machine (SVM) and random forest (RF) were developed and applied for the classification of each type of samples including gold, meat, and gemstone through Python programming language as it is open-source, and it offers a comprehensive list of libraries and packages unlike other softwares that include licensing and subscription fees. Principal component analysis (PCA) was employed on the three samples to visualize the data. The scree plot of the PCA technique has been generated, which has expressed that the first PC scores of golds, meats and gemstones were 90.1%, 96.1%, 96.4% accordingly. It has been concluded that, the Polynomial kernel from the SVM classifier has worked best for metallic samples with the classification accuracy of 89.7% while the organic samples have been well classified through the RBF kernel by SVM classifier which expressed the accuracy of 90.0%. The Polynomial kernel SVM, the RBF kernel SVM along with the RF algorithms are preferred to be utilized on the geological samples as they have shown the highest classification accuracy on the samples which were a perfect 100%.