Data Mining as a Tool to Infer Chicken Carcass and Meat Cut Quality from Autochthonous Genotypes.

The present research aims to develop a carcass quality characterization methodology for minority chicken populations. The clustering patterns described across local chicken genotypes by the meat cuts from the carcass were evaluated via a comprehensive meta-analysis of ninety-one research documents published over the last 20 years. These documents characterized the meat quality of native chicken breeds. After the evaluation of their contents, thirty-nine variables were identified. Variables were sorted into eight clusters as follows; weight-related traits, water-holding capacity, colour-related traits, histological properties, texture-related traits, pH, content of flavour-related nucleotides, and gross nutrients. Multicollinearity analyses (VIF ≤ 5) were run to discard redundancies. Chicken sex, firmness, chewiness, L* meat 72 h post-mortem, a* meat 72 h post-mortem, b* meat 72 h post-mortem, and pH 72 h post-mortem were deemed redundant and discarded from the study. Data-mining chi-squared automatic interaction detection (CHAID)-based algorithms were used to develop a decision-tree-validated tool. Certain variables such as carcass/cut weight, pH, carcass yield, slaughter age, protein, cold weight, and L* meat reported a high explanatory potential. These outcomes act as a reference guide to be followed when designing studies of carcass quality-related traits in local native breeds and market commercialization strategies.