Applications of Dynamic Pattern Recognition Methods

Abstract

In order to demonstrate the practical relevance of the new method for dynamic fuzzy clustering developed in this book, two application examples are considered in this chapter. The first example taken from credit industry and presented in Section 6.1 is concerned with the problem of bank customer segmentation based on customers’ behavioural data. After a description of the credit data of bank customers and the formulation of the goals of the analysis, two types of customer segmentations based in the first case on the whole temporal history covering two years and in the second case on a partial temporal history of half a year will be carried out. The clustering results obtained in the first case represent the structure within the customer portfolio related to long-term payment behaviour whereas the results generated in the second case provide customer segments based on short-term behaviour and the information about temporal changes in customer behaviour. This section contains a detailed description of the customer segments obtained during both types of analysis, an evaluation of the quality of the fuzzy partitions and a comparison of the different clustering results. The second application example, presented in Section 6.2, is related to the analysis of data traffic in computer networks and allows the optimisation of the network load based on on-line monitoring and dynamic recognition of typical states of data traffic. Dynamic fuzzy classifier design and classification will be performed based on pointwise as well as structural similarity measures for trajectories and the clustering results will be compared and evaluated.