[Local Regression Algorithm Based on Net Analyte Signal and Its Application in Near Infrared Spectral Analysis]

Area of Science:

• Analytical Chemistry
• Chemometrics

Context:

• Spectral quantitative analysis often faces challenges with sample variability and nonlinear relationships between spectral data and sample properties.
• Accurate analysis of complex samples like meat requires robust chemometric methods.

Purpose:

• To develop a novel local regression algorithm that addresses the limitations of traditional spectral quantitative analysis.
• To improve the prediction precision and reduce the model complexity in analyzing near-infrared spectra of meat samples.

Summary:

• The proposed algorithm utilizes Net Analyte Signal (NAS) to establish a similarity index based on Euclidean distance.
• Local calibration sets are dynamically selected for each unknown sample, enabling the development of localized Partial Least Squares (PLS) regression models.
• This approach was validated using near-infrared spectra of meat samples, demonstrating superior performance over global PLS and conventional local regression methods.

Impact:

• The developed local regression algorithm offers a more accurate and efficient method for spectral quantitative analysis.
• This technique has the potential to enhance quality control and analysis in food science and other industries relying on spectral data.