Determination of gasoline octane number using Raman spectroscopy and least squares support vector machines

This paper presents a novel algorithm to predict gasoline octane number with Raman spectroscopy. The algorithm is based on Least Squares Support Vector Machines (LS-SVM). It uses the 2-D digital filters, De-trending and Standard Normal Variate (SNV) transformation to preprocess the original Raman data, and introduces cluster analysis to select a training set from the processed Raman data. Finally, it utilizes the LS-SVM to build the prediction model of gasoline octane number based on the training set. Experimental results show that the proposed algorithm can obtain better prediction performance than regular algorithms such as Partial Least Squares.