Fluorescence Spectra Detection Based On Optimized Support Vector Machine

People in today's society attach great importance to the problem of environme nta l pollution and food safety.Using new methods to study the characteristics of environme nta l pollutants and food additives will not only help improve the quality of the environment and ensure food safety,but also provide a possible experimental method for the study of other substances.According to the mechanism of fluorescence excitation,both benzo-type substances and potassium sorbate preservatives have strong fluorescence characterist ics.This paper uses support vector machine combined with some intelligent algorithms as a model to establish an optimized support vector sorter that can realize the type identifica t io n of three types of benzo-type mixtures.,and establish an optimized support vector regression machine that can predict the concentration of potassium sorbate in orange juice,so that the purpose of learning qualitative and quantitative analysis methods can be achieved.The main research contents of this article are as follows:(1)Under the theoretical background of the fluorescence excitation mechanism,the general framework of the FS920 fluorescence spectrometer and the selection of the main part are proposed.Fluorescence experiments are carried out on the three kinds of benzotype mixtures and their fluorescence characteristics are analyzed.The appropriate fluorescence spectrum data of the three types of substances is selected and the training sample set and prediction sample set are obtained.By analyzing the fluoresce nce characteristics of potassium sorbate in orange juice,suitable spectral data is selected and the training sample set and test sample set are constructed.(2)Based on the classification principle of support vector machine,the improved chicken swarm optimization algorithm(ICSO)is used to optimize the support vector machine.The samples of benzo-substances is trained with the established ICSO-SVM model,and then the prediction results for test set are compared with the CSO-SVM,GASVM and PSO-SVM models.The results show that the ICSO-SVM model is the best in the fitness function during the training and the accuracy of the prediction results.(3)The principle of support vector regression machine(SVR)is analyzed,and then the support vector machine model optimized by cuckoo search algorithm(CS)is built.The CSSVM model is trained with the training sample set of potassium sorbate preservative,and the GA-SVM model and PSO-SVM model are established at the same time.Through the comparison of its fitness function,recovery rate and mean square error,it is finally shown that the CS-SVM model is optimal and it is validated that the advantages of the CS-SVM model.