Detection of Chicken Meat Freshness Using NIR spectroscopy and Machine Learning Algorithms

Resumen

Chicken meat freshness is a critical quality and safety indicator that requires rapid and non-destructive detection methods. This study aimed to develop a spectral classification model to identify chicken freshness using NIR spectroscopy and machine learning algorithms. A total of 180 spectra (1100–2495 nm) were collected and labeled as “fresh” (days 1–3) or “spoiled” (days 4–5). After Savitzky–Golay smoothing and SNV correction, five models (SVM, LASSO, Ridge, Elastic Net, Random Forest) were trained using stratified cross-validation. The LASSO model achieved the best performance with 97.2% accuracy and AUC = 0.977, with no false negatives in the spoiled class, followed by SVM (94.4%). Results demonstrate the effectiveness of regularized linear architectures for detecting chemical changes associated with spoilage. This approach provides a rapid, non-invasive, and accurate tool for freshness monitoring in poultry production chains.