Classification of olive oils according to their cultivars based on second-order data using LC-DAD

Second-order data acquired using liquid chromatography coupled to a diode array detector were used to classify extra virgin olive oils samples according to their cultivars. The chromatographic fingerprints from the epoxidised fraction were obtained using normal-phase liquid chromatography. To reduce the data matrices two strategies were employed: (1) multivariate curve resolution-alternating least squares (MCR-ALS) and (2) a new strategy proposed in this work based on the fusion of the mean data profiles in both spectral and time domains. Several conventional chemometric tools were then applied to both raw and reduced data: principal component analysis (PCA), partial least-squares-discriminant analysis (PLS-DA), soft independent modelling of class analogies (SIMCA) and n-way partial least-squares-discriminant analysis (NPLS-DA). Furthermore, an emergent multivariate classification method known as random forest (RF) has been first applied to second-order data. It was shown that RF is more efficient than conventional tools. Indeed, the obtained sensibility, specificity and accuracy are 1.00, 0.92 and 0.95 respectively; these performance metrics are significantly better than the values found for the other methods.