Rapid Authentication and Quality Assessment of Fish oil Using Near Infrared Spectroscopy

A near-infrared (NIR) spectroscopy-based framework was developed to efficiently evaluate fish oil quality. The workflow focuses on three main objectives: quantification of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), detection of adulteration, and classification of purity. For EPA/DHA quantification, spectra were preprocessed, and partial least squares regression (PLSR) was compared with support vector regression, where PLSR achieved higher accuracy and more stable performance. For adulteration quantification, pure fish oil was spiked with rapeseed, soybean, corn, and palm oils at concentrations ranging from 0% to 50% (w/w). Linear discriminant analysis of NIR spectra revealed distinct separation between pure and highly adulterated samples, with partial overlap at intermediate levels. PLSR models predicted adulterant percentages for all four oils with prediction R² values ≥ 0.98 and low error. For purity classification, random forest models based on NIR spectra showed strong performance, and integration of selected gas chromatography variables through mid-level data fusion further improved accuracy to 91.7%. Overall, this NIR-based workflow provides a rapid, nondestructive, and reliable tool for label verification, supplier management, and routine quality assurance in fish oil products.