The feature of gasoline Raman spectra which were used to study the quantitative analysis of the research octane number (RON) were extracted for the first time using backward interval partial least squares (BiPLS). In the experiment, the sample set partitioning based on joint x-y distances (SPXY) method was used to divide the training set, the cross validation set and the test set. And the robust regression algorithm was used to remove the abnormal sample. The partial least squares model was established using feature selected by the BiPLS algorithm. Compared with the model without feature selection, it was shown that the backward interval partial least squares algorithm could reduce the input dimension by 50.00%, and the root mean square error of cross validation(RMSECV) by 18.92% and the root mean square error of prediction (RMSEP) by 13. 86%. The backward interval partial least squares algorithm can effectively extract the feature from gasoline Raman spectrum,reduce the model complexity, and improve the prediction accuracy of the model,and has great application prospect in the quantitative analysis of research octane number.