Qualitative radial basis function networks applied to financial credit risk prediction

A methodology to use RBF when the descriptors of the patterns are given by means of ordinal variables is developed. A qualitative distance is constructed over the discrete structure of absolute orders of magnitude spaces. The aim is to capture the remoteness between the components of the patterns by locating labels with respect to extreme magnitudes. An application to a financial problem of the learning method described is given and permits a comparison of results obtained from a qualitative treatment with those from a quantitative treatment.