Recognition of polycyclic aromatic hydrocarbons using fluorescence spectrometry combined with bird swarm algorithm optimization support vector machine
Shutao Wang,Shiyu Liu,Xiange Che,Zhifang Wang,Jingkun Zhang,Deming Kong
DOI: https://doi.org/10.1016/j.saa.2019.117404
2020-01-01
Abstract:<p>Polycyclic aromatic hydrocarbons (PAHs), known as a widespread toxic pollutants in aquatic environments, have caused enormous harm to human society and even the earth's ecology. Therefore, it is necessary to identify PAHs pollutants accurately and efficiently. In this work, the binary mixed solvents Acenaphthylene and Fluorene (ANP-FLU), Acenaphthylene and Naphthalene (ANP-NAP), FLU-NAP representing typical PAHs mixtures in aqueous solution were identified by using three-dimensional fluorescence spectroscopy and machine learning intelligent algorithm. The fluorescence spectroscopy was used to analyze the similarity and difference of ANP, FLU, NAP and the mixtures of two above compounds. What's more, bird swarm algorithm optimization support vector machine (BSA-SVM), introduced as a new method, was proposed to identify PAHs. In order to verify the accuracy of the BSA-SVM algorithm, the BSA-SVM, particle swarm optimization support vector machine (PSO-SVM), genetic optimization support vector machine (GA-SVM) and SVM algorithms were test by processing the same spectral data. The test set classification accuracy of BSA-SVM can reach 100%, which was higher than that of PSO-SVM, GA-SVM and SVM. Moreover, with the exception of the original SVM model, the training speed of BSA-SVM was the fastest among the three optimization algorithms. The satisfying results demonstrated that the BSA-SVM was more suitable for qualitative analysis of PAHs.</p>