<![CDATA[This study investigates the application of near-infrared spectroscopy (NIR) within the wavelength range of 1350–2550 nm to predict key quality parameters of beef, specifically focusing on tenderloin cuts. The quality indicators assessed include drip loss, color, pH, moisture content, storage duration, and total plate count (TPC) as a measure of microbial load. Predictive modeling was conducted using three machine learning algorithms: Partial Least Squares (PLS), Support Vector Regression (SVR), and Random Forest Regressor (RFR). To enhance model accuracy, a semi-automated preprocessing pipeline was employed utilizing the Nippy library. This library integrates several spectral preprocessing techniques including Savitzky-Golay filtering, Standard Normal Variate (SNV), Robust Normal Variate (RNV), Local Standard Normal Variate (LSNV), as well as clipping, resampling, baseline correction, and smoothing. Among the models developed using raw spectral data, the RFR model exhibited the highest performance, achieving coefficient of determination (R²) values of 0.82 for drip loss, 0.65 for color, 0.67 for pH, 0.61 for moisture content, 0.81 for storage duration, and 0.76 for TPC. Post preprocessing, the predictive accuracy improved significantly with R² values increasing to 0.89, 0.82, 0.87, 0.85, 0.91, and 0.90 respectively for the same parameters. These findings underscore the potential of combining advanced machine learning techniques with robust preprocessing methods to enhance the non-destructive, rapid assessment of beef quality parameters. This approach offers a promising tool for quality control in the meat processing industry, facilitating more efficient and accurate monitoring of product standards.]]>
