Effizientes K-Clustering: K-Centroids Clustering auf Basis von hierarchisch aggregierten Daten

In the world of Knowledge Discovery in Databases (KDD) exists a set of problems which can hardly be solved by following the commonly known process, for example if one tries to re-run an algorithm with different parameters. This thesis identifies such problems and justifies why a new approach in KDD is required. This thesis introduces the new approach "Sequential Data Mining" and locates it in the commonly known process. The idea behind "Sequential Data Mining" does not deal with a single appliance of an algorithm to a dataset, but a multiple analysis of data, thereby subsequent runs benefit from the results of previous. A new algorithm which follows this concept and is able to cope with mentioned problems by generating efficient proper results is introduced. The method is able to produce the result of a K-Clustering algorithm by using an apriori aggregated representation of data, hence the title of this thesis is "Efficient K-Clustering". The new method will be differentiated of other well known approaches and compared to them. Further the potential of the new algorithm will be validated using different tests.