Geographical origin traceability of kiwifruit products using stable isotope and multi-element analysis with multivariate modeling: Feature extraction, selection of model and variable, and discrimination.

The mislabeling of kiwifruit origin frequently disturbs market competition and governmental supervision, significantly undermines brand reputation and consumer rights. In this work, a total of 370 kiwifruits from 8 different countries in global were collected, and 6 stable isotope ratios (SIRs), 10 mineral elements (MEs), and 16 rare earth elements (REEs) were determined for origin traceability study. One-way analysis of variance (ANOVA) showed that regional differences of 32 variables are at significant level ( value =0.00). Supervised methods, partial least squares-discriminant analysis (PLS-DA) and its derivative algorithm (OPLS-DA), linear discriminant analysis (LDA), enhanced identification performance and finally elevated the accuracies to 100 % for all kiwifruit origins. Lu, Tb, Eu, Ho, Pm, Y, δS, δH, δN, Mg, Se were main contributive variables for LDA modeling (AUC value >0.5). A blind test was conducted using 63 samples randomly selected from Chinese market. The predicted result indicated a significantly high probability of origin mislabeling of imported kiwifruit products, with percentages ranging from 30.0 % to 90.0 %. This study may provide technical supports for combating origin mislabeling conduct, and ensuring food authenticity of kiwifruit in global trade.