Different parts of dangyuja (Citrus grandis Osbeck) fruits at different maturation stages were classified using a (1)H NMR-based metabolomic technique. Principal components analysis allowed the clear separation of fractions extracted with 50% methanol of different parts of dangyuja fruits at different maturation stages by combining principal components PC1 and PC2, which together accounted for 80.4% of the variance. A loading-plot analysis revealed that sucrose, glucose, oxaloacetic acid and citric acid were dominant in mature flesh, while naringin, tyramine, proline and alanine were dominant in immature fruit samples. Projections to latent structures using a partial least squares (PLS) model were used to predict the free-radical scavenging activities (FRSA) of dangyuja fruit extracts based on their (1)H NMR spectra. The present study suggests the usefulness of combining (1)H NMR spectroscopy with multivariate statistical analysis for discriminating dangyuja fruit samples, and predicting the FRSA of different parts of dangyuja fruit samples at different stages of maturation.