Title : Rapid detection of milk adulteration using NIR spectroscopy and machine learning techniques
Abstract:
Milk adulteration remains a significant challenge to food safety and quality assurance. In this study, a novel Near-Infrared (NIR) spectroscopy and Machine Learning (ML)-based framework was developed for the rapid detection of common adulterants in raw cow milk. The acquired spectral data were preprocessed using Savitzky-Golay filtering, Multiplicative Scatter Correction (MSC), and Standard Normal Variate (SNV) techniques to improve spectral quality and minimize scattering effects. Feature extraction was performed from characteristic spectral regions following derivative and SNV transformations. Both regression and classification approaches were employed to evaluate the performance of the proposed system. Among the evaluated models, Support Vector Regression (SVR) demonstrated strong predictive capability for estimating adulterants concentration in milk, while Support Vector Machine (SVM) achieved highly accurate classification of adulterated samples. The proposed approach provides a rapid, reagent-free, and non-destructive solution with strong potential for integration into portable devices for real-time milk quality monitoring and food safety applications.
