Abstract
Morphometric traits provide valuable insights into poultry breed characterization, yet their predictive utility across developmental stages remains unexplored. This study aimed to categorise three poultry strains (PB1, PB2, and Punjab Brown) using morphometric data across growth phases, employing statistical and machine learning methodologies. One-way ANOVA, correlation matrices, and dimensionality reduction methods like Principal component analysis (PCA), Kernel PCA, and Linear Discriminant Analysis (LDA) were used to measure and analyse 29 morphometric traits. Logistic regression (LR) models classified strains on reduced-dimensional datasets. Significant (p 0.85) were observed in girth and skeletal dimensions, whereas early expressed traits such as hatch weight (BW0) showed low correlations (r < 0.30) with other variables across different strains, indicating limited predictive value. PCA retained 85.6% of variance within the first five components, and LDA enabled distinct strain separation. Among classification pipelines, LDA + logistic regression achieved the highest performance (accuracy, precision, and recall = 1.00), followed by KPCA + LR and PCA + LR. Integrating morphometric profiling with supervised learning effectively differentiates poultry strains and enhances classification accuracy.