Advanced Chemometric and Machine Learning Techniques in NIR Spectroscopy for Freshness Class Prediction of Chicken Eggs

Abstract

The quality of eggs deteriorates during storage, making freshness monitoring crucial in the egg industry. This study explores the use of Near-Infrared (NIR) spectroscopy, a rapid and non-destructive method, combined with chemometric techniques to assess and classify egg freshness effectively. A total of 660 eggs were stored at controlled temperatures (20°C and 30°C) and observed for 21 days, with spectral data collected across a range of 902-1810 nm at 4 nm intervals. The freshness was analyzed in relation to Haugh Units (HU) and days of storage. To enhance data quality, preprocessing methods such as Multiplicative Scatter Correction (MSC), Standard Normal Variate (SNV), and Savitzky-Golay smoothing were applied. Dimensionality reduction through Principal Component Analysis (PCA) helped streamline data, while predictive models, including Partial Least Squares Regression (PLSR) and Support Vector Machine Regression (SVM-R), were developed to estimate HU values accurately. The study achieved an R² of 0.997 in calibration, indicating strong predictive power, especially with SVM-R. Additionally, classification models Partial Least Squares Discriminant Analysis (PLS-DA) and Support Vector Machine Classification (SVM-C)—achieved up to 99.20% accuracy in distinguishing between different freshness levels. These findings underscore NIR spectroscopy's potential as a reliable tool for real-time quality monitoring, offering efficiency, accuracy, and non-destructive testing for the egg industry. This approach could serve as a valuable alternative to conventional freshness assessments, providing precise insights into egg quality during storage, thereby ensuring better quality control and consumer satisfaction in the food industry.