Fluorescence spectroscopy offers the possibility to make qualitative and quantitative analysis of crude oils. Its sensitivity and selectivity permit its use for oils exploration and production. The study describes a novel approach for identification and discrimination of oils by chemometric treatment of synchronous ultraviolet fluorescence (SUVF) spectra which are usually studied by indexes method. Clusters of samples are obtained according to their geographic origin on a 3D plane using SUVF indexes and in the space of principal components for PCA method. Evaluation of oils composition is performed with MCR-ALS algorithm by extracting spectra of families representing aromatic structures. A PLS calibration, built with spectra intensities as input and oils origin as output allows to identify oils chosen for the prediction. Results show that mathematical exploration of full spectral region leads rapidly to classify and predict some oils properties according to their geochemistry. (c) 2006 Elsevier Ltd. All rights reserved.