Geographic origin discrimination of red swamp crayfish Procambarus clarkii from different Chinese regions using mineral element analysis assisted by machine learning techniques

shuyan Bai,dongli Qin,zhongxiang Chen,song Wu,shizhan Tang,peng Wang

DOI: https://doi.org/10.1016/j.foodcont.2022.109047

IF: 6

2022-04-01

Food Control

Abstract:To explore the application prospects of mineral element analysis assisted by machine learning in the geographical origin traceability of red swamp crayfish (Procambarus clarkii) and establish an origin traceability model for it, we used inductively coupled plasma mass spectrometry (ICP-MS) to determine the content of 24 mineral elements, including Cu, Zn, Ga, As, Se, Rb, Sr, and Mo, in red swamp crayfish muscle tissues from Anhui, Hunan, and Hubei provinces of China. The ten characteristic elements, screened by the principal component analysis (PCA), were incorporated into the linear discriminant analysis (LDA) to establish an origin discriminant model of red swamp crayfish. We have verified the feasibility of the mineral element fingerprinting to trace the geographical origin of red swamp crayfish in three regions of China. To improve the discrimination performance, three machine learning classification algorithms, the support vector machine (SVM), the feed-forward neural networks (FNN), and the back-propagation neural network (BPNN), were combined to classify the red swamp crayfish of different origins based on mineral element content. The discriminative accuracy of the training models of both FNN and BPNN reached 100%, and their cross-validation accuracy were 91.88% and 93.13%, respectively, with the area under the receiver operating characteristic (ROC) curve (AUC) values close to 1. Thus, the overall performance of the FNN and BPNN was better than that of SVM. Taken together, our results showed that mineral elements combined with neural networks showed great advantages and potential in geographic origin traceability of red swamp crayfish.

food science & technology