Multi-attribute classification of credit cardholders: multiset approach

Abstract

A new technique for classifying real objects, which is described by many numerical and verbal attributes, is considered in this paper. In this method, general classification rules are generated by a proper approximation of the given set of diverse individual sorting rules. The classification technique is based on the theory of multiset metric spaces, where multisets represent both properties of multi-attribute objects and sorting rules. The proposed technique allows us to operate with contradictory and arbitrary-size data. The presented approach was applied to the cardholder credibility categorisation.