In the course of the last years enormous quantities of data were entered in the enterprise data bases, which actual information content often remains unused. In the future the competitiveness of an enterprise is influenced crucially, if it at pending decisions or planning activities is able to detect and consider the actual "value" or the information content of these data. Therefore, progressive enterprises lately use modern data analysis techniques, which try to penetrate the available data jungle to extract the urgently necessary information. With the help of so-called Data Mining procedures often completely new connections or interactions between data are obtained, which usually are unknown and also not evident at first sight.

In the area of geographical information systems (GIS) data analyses - if at all - are executed still manually and partly still also aborted without result. The principal reason for this situation is to be looked up in the fact that commercially available Data Mining tools are not able to analyze the complex and heterogeneous volume of data of geographical information systems. The few known GIS specific Data Mining procedures are still tested at present in the academic area and concentrate themselves primarily on revealing connections between the spatial position and actual pertinent data of objects. Correlations with consideration of network topologies, which form the heart of many volumes of geographical data, are so far not detected by any existing procedure.

In the context of this thesis (diploma) the theoretical bases for a cluster procedure on network structures were both acquired and developed. This procedure was prototypically implemented in a Data Mining tool, as well as tested finally in a case study.

This dissertation is divided beside the short introduction in the first section and the bibliography into six further sections. In the second section the most important bases of existing conventional and GIS specific Data Mining procedures are presented, so that there is sufficient information for the reader to understand the following remarks. In the third section an existing GIS specific procedure, which combines point objects into cluster concerning their distance, will then transfer to network structures. In addition first the necessary theoretical bases are compiled, in order to define afterwards a cluster procedure on network structures, which will combines either node or edge objects into cluster, on base of a configurable weighting function with consideration of the network topology. In the fourth section the development process to the implemented Data Mining tool is the center of interest. Here are important notes to the implementation of the central architecture components described. In the fifth section (user guide) the most important steps are presented for the operation of the prototype. In the process of the dissertation work the prototype or rather the developed cluster procedure was successfully tested at a public utility company on the basis of a real problem definition. The sixth section shows an overview of this case study. A summary and a concluding evaluation of the entire procedure are at the end of the work.