clustering; object-oriented modeling; BETA

In this paper we describe results achieved by combination of the hierarchical and partitional clustering. Proper design and object-oriented approach bring new possibilities of the various combinations of the strategies in numerical clustering. One of these possibilities enables to set out with the hierarchical clustering and at some thoughtful level continues with the partitional clustering by adding new clusters. In this way we obtain numerical characteristics for the newly added clusters that may be used for making diagnoses. Actual state of the hierarchical clustering is stored and retrieved by using persistent store and persistent objects. Described approach was tested on medical data. The whole application is developed using the Mjolner BETA System and the BETA language for its rich facilities of the conceptual modeling.

In this paper we describe results achieved by combination of the hierarchical and partitional clustering. Proper design and object-oriented approach bring new possibilities of the various combinations of the strategies in numerical clustering. One of these possibilities enables to set out with the hierarchical clustering and at some thoughtful level continues with the partitional clustering by adding new clusters. In this way we obtain numerical characteristics for the newly added clusters that may be used for making diagnoses. Actual state of the hierarchical clustering is stored and retrieved by using persistent store and persistent objects. Described approach was tested on medical data. The whole application is developed using the Mjolner BETA System and the BETA language for its rich facilities of the conceptual modeling.