Comparing data driven soft independent class analogy (DD-SIMCA) and one class partial least square (OC-PLS) to authenticate sacha inchi (Plukenetia volubilis L.) oil using portable NIR spectrometer

Abstract

Sacha Inchi (SI) oil is valuable due to high content of omega-3 and omega-6, which are important in human diet, thus being susceptible to adulteration. In this work, a low-cost and portable Near Infrared (NIR) spectrometer was used to authenticate pure sacha inchi oil from adulterated one with cheaper oils. Principal component analysis (PCA) was applied as screening technique and showed a good class separation between pure SI oil and blends, with significant impact of unsaturated fatty acids present in SI oil. Data Driven Soft Independent Class Analogy (DD-SIMCA) provided better results when compared to One Class Partial Least Square (OC-PLS) for data analysis. DD-SIMCA was implemented in rigorous approach, and it reached Sensitivity (SEN) = 89.8 % and Specificity (SPE) = 93 % using α = 0.05 and 4 PCs, while ordinary OC-PLS required 10 LVs to reach SEN = 87.8 % and SPE = 84.3 %, performing better than Gaussian radial basis function (GRBF-OCPLS) and partial robust M-regression (PRM) approaches. Results indicate that low-cost NIR spectrometer coupled to one-class classifiers could be used to authenticate pure SI oil.