A strategy to significantly improve the classification accuracy of LIBS data: application for the determination of heavy metals in Tegillarca granosa

Abstract

Tegillarca granosa, as a popular seafood among consumers, is easily susceptible to pollution from heavy metals. Thus, it is essential to develop a rapid detection method for Tegillarca granosa. For this issue, five categories of Tegillarca granosa samples consisting of a healthy group; Zn, Pb, and Cd polluted groups; and a mixed pollution group of all three metals were used to detect heavy metal pollution by combining laser-induced breakdown spectrometry (LIBS) and the newly proposed linear regression classification-sum of rank difference (LRC-SRD) algorithm. As the comparison models, least regression classification (LRC), support vector machine (SVM), and k-nearest neighbor (KNN) and linear discriminant analysis were also utilized. Satisfactory accuracy (0.93) was obtained by LRC-SRD model and which performs better than other models. This demonstrated that LIBS coupled with LRC-SRD is an efficient framework for Tegillarca granosa heavy metal detection and provides an alternative to replace traditional methods.