Fig 1. A diagram representing the processes of building the prediction model for traceability.

The prescreening process for selecting the SNP markers consists of two major steps: retrieval of common SNPs for five pig breeds and selection of SNP markers based on geographical distribution (farm location). Farms were filtered by the kinship coefficient mean and four subsets were generated. The feature selection process for removing redundant features was performed using two approaches (detailed descriptions of these techniques are provided in the manuscript) and three classifiers. Using the selected features, classification performance was evaluated based on three factors (classification accuracy, balanced accuracy, and ROC curves).