A non-standard-substance pesticide residue qualitative analysis method based on SVM

Abstract

Currently, most pesticide residue analysis methods are based on mass spectrometry instruments, such as Gas Chromatography-Mass Spectrometry (GC-MS) analyzers. However, using these methods, user has to send the chemical standard-substances (pure pesticides) and the samples needed checking simultaneously in order to detect the pesticide residues in foods and agricultural products. Support Vector Machines (SVM) is a statistical learning method based on small-sample set. It has the advantages in solving the small-sample set, nonlinear and high dimension problems and is widely used in classification. A non-standard-substance pesticide residue qualitative analysis method (NSS-QAM) is presented in this paper. NSS-QAM transforms qualitative analysis problem into a problem of classification of pesticide residues based on SVM and existing official standards for pesticide residue detection and experiment results. NSS-QAM is divided into two steps. The first one is to let SVM learning according to characteristic parameters of pesticide residue from existing standards and experiment results and get classification model. This step is executed by offline and applies one-versus-one multi-class classification SVM. The second step is to use the model to classify detecting data from GC-MS analyzer and implement qualitative analysis for pesticide residue without using chemical standard-substance. NSS-QAM has been experimented with 1500 samples from 50 pesticides. The classification results demonstrate that NSS-QAM is an effective qualitative analyzing method.