Fig. 1: Schematic workflow of machine learning-assisted Raman spectroscopy for non-destructive quality assessment of crude palm oil.

This figure illustrates the complete workflow for the non-destructive quality assessment of crude palm oil using machine learning-assisted Raman spectroscopy. The process begins with sample acquisition, followed by data collection using standard reference methods, including the determination of iodine (IV) and peroxide (PV) values as per AOAC protocols. The Raman spectral data (500–2000 cm⁻¹) is then preprocessed using various techniques, such as Savitzky-Golay smoothing, wavelet denoising, and derivative transformations (1st and 2nd derivatives). The next step involves variable selection using methods such as CARS, UVE, and GA, which are used to optimize the data. Subsequently, regression modeling is applied using various machine learning algorithms, including PLS, CARS-PLS, UVE-PLS, GA-PLS, SVM, and Random Forest, to predict the quality parameters of crude palm oil. The regression results section displays the correlation between the predicted values and reference measurements, with a strong predictive ability indicated by the high correlation coefficients (Rc and Rp).