An ensemble of classifiers for the diagnosis of erythemato-squamous diseases

A new ensemble of support vector machines (SVM) based on random subspace (RS) and feature selection is developed and applied to the problem of differential diagnosis of erythemato-squamous diseases. Each classifier has a ''favourite'' class. To find the feature subset for the classifier D"i with ''favourite'' class w"i, we calculate the best features to discriminate this class (w"i) from all the other classes. Our results improved the average predictive accuracy obtained by a ''stand-alone'' SVM or by a RS ensemble of SVM.