Study on Apostichopus Japonicus Origin Identification Online System Based on PCA-SVM
Wu Peng,Li Ying,Liu Yu,Fu Jin-yu,Li Ya-fang,Ran Ming-qu,Zhao Xin-da
DOI: https://doi.org/10.3964/j.issn.1000-0593(2019)05-1604-06
2019-01-01
Spectroscopy and spectral analysis
Abstract:Apostichopus japonicus, an important part of the mariculture, is a fishery resources with extremely high economic value. Therefore, it is of great practical significance for the mariculture to study a flexible, stable and efficient method to identify the origin information of apostichopus japonicas. There are three main aquaculture methods for the apostichopus japonicas, including bottom sowing culture, captive culture and raft culture. Different aquaculture methods are used in apostichopus japonicus of different producing areas, and there also exist great differences in the nutritional value, medicinal value and economical value from different producing areas. The compositions of primary producers vary from one provenance to another, and the characteristics of fatty acids in different algae and plankton as primary producers are also different. Through the transmission of the food chain, apostichopus japonicus from different producing areas have different fatty acid characteristics. Gas chromatographic fingerprint is a fast and accurate traceability technology for food origin. Carbon stable isotope ratio mass spectrometry can not only identify origin but also distinguish the nutritional value of food. Samples of apostichopus japonicus were collected from nine representative producing areas, and total lipid data were extracted by using the Folch method. Then determined the data of fatty acid kinds and relative content through gas chromatography. Finally, stable isotope ratio mass spectrometer was used to determine the data of fatty acids carbon stable isotope compositions. One-factor analysis of variance (ANOVA) was used to test the significance of the data of fatty acid relative content and fatty acids carbon stable isotope compositions, and then selected 17 kinds of fatty acid data as inputs for the two models. The principal component analysis(PCA) method can reduce the dimensions of the data, and aggregate the origin characteristics of different fatty acids to improve the accuracy of origin identification model. Support vector machine (SVM) is a classification algorithm that aims to minimize structural risk and has good ability of generalization. The results indicated that there were significant differences in the fatty acids relative content and fatty acids carbon stable isotope compositions data of apostichopus japonicus from different producing areas. After the principal component analysis, the clustering characteristics of the fatty acids data were more obvious. With the cross-validation method, the first six principal components were determined as inputs of the two support vector machine classifiers. Applied the particle swarm optimization improved based on genetic crossover factor(GPSO) , and the average accuracy of 100 cross-validation results of different K values of the particle was used as the fitness to find optimal parameter combinations of the SVM classifier model. Finally, the optimal parameter combinations of fatty acids relative content model were sigma = 6. 247 599 and C = 14. 313 042, and average accuracy was 79. 49%; the optimal parameter combinations of fatty acids carbon stable isotope compositions model were sigma=7. 626 194 and C = 2. 193 410, and average accuracy was 98. 33%. Compared with the results of cross-validation, the fatty acids carbon stable isotope compositions origin identification model has higher accuracy and stronger ability of generalization. When origin identification results of two models were inconsistent, the results of fatty acids carbon stable isotope compositions model were used. The laboratory detection was integrated with Internet technology to build an apostichopus japonicus origin identification online system. The Integrated model of "Internet+origin identification" has achieved to provide a scientific basis and technical support for the further studies on origin identification of food.