A least squares SVM algorithm for NIR gasoline octane number prediction

This paper presents a novel algorithm, based on least squares support vector machines (LS-SVM), to predict gasoline octane number with near-infrared (NIR) spectroscopy. This algorithm not only has the same high generalization performance and global optimal solution as standard SVM, but also needs less computing time, which is necessary to on-line application. Experimental results show that the proposed algorithm can obtain better prediction performance than regular algorithms such as multivariate linear regression and partial least squares.