Classification of Raw Açai (Euterpe oleracea Mart.) Fruits from Upland and Lowland Environments Using Portable Near-Infrared Spectroscopy

Near-infrared (NIR) spectroscopy combined with chemometrics was applied to classify Euterpe oleracea (açai) fruits according to their soil environment. Raw fruit samples were analyzed using a handheld NIR spectrometer. Compositional differences between soil environments were confirmed by reference data and the most discriminative NIR regions were related to moisture, phenolics, and lipids. Discriminant models outperformed one-class models. Partial Least Squares Discriminant Analysis (PLS-DA) and Support Vector Machines Discriminant Analysis (SVM-DA) provided perfect or near-perfect classifications (F1 = 0.97–1.00), confirming the generalization ability of latent-variable and kernel-based strategies. Variable selection improved model interpretability without compromising accuracy. One-class models, Soft Independent Modelling by Class Analogy (SIMCA and DD-SIMCA) and One-class partial least squares (OCPLS), exhibited broader variability for upland fruits, reflecting greater intra-class chemical diversity. Handheld NIR combined with multivariate classification provides a rapid, reliable, and sustainable analytical approach for discriminating açai fruits by soil environment, offering a promising tool for origin verification and quality control in the açai production chain.