Computer-aided interpretation of steroid mass spectra by pattern recognition methods: Part III. Computation of Binary Classifiers by Linear Regression

Abstract

The method of linear regression analysis is used to compute binary linear classifiers which can recognize 17 chemical structures of steroids from given low-resolution mass spectra. Different ways of spectral preprocessing and feature selection are compared. The best classification results are obtained with spectra which are normalized to local ion current and with feature selection based on maximum Fisher ratio. An average predictive ability of 84% was achieved for a spectral sample which was not used for the computation of the classifiers.