Random forest as a potential multivariate method for near-infrared (NIR) spectroscopic analysis of complex mixture samples: Gasoline and naphtha

Abstract

Random forest (RF) has been demonstrated as a potential multivariate method for near-infrared (NIR)spectroscopic analysis of petroleum-driven products, highly complex mixtures of diverse hydrocarbons.For the study, a NIR dataset of gasoline samples and two separate NIR datasets of naphtha samples wereprepared. These samples were carefully prepared over a long period to maximize compositional variationin each dataset. Partial least squares (PLS), the most widely adopted method in multivariate analysis, andRF were used to determine research octane numbers (RONs) of gasoline samples, and total paraffin, totalnaphthene and total aromatic concentrations of naphtha samples. The resulting accuracies of quantitative analysisfor these samples were generally improved when RF was used. In addition, chance for overfitting of a model, whichwould occur occasionally in PLS modeling, was substantially lessened or possibly eliminated by the use of RF. On thecontrary, in the case of RF, a calibration dataset composed of samples with narrow interval in property or concentrationvariation was required to improve the accuracy. Consequently, RF could be a useful multivariate methodto analyze NIR as well as other spectroscopic data acquired from petroleum refining products, when properlyutilized.