Using a Clustering Algorithm for Domain Related Ontology Construction

Fisher's clustering algorithm is exploited to build a cluster hierarchy. Then this methodology is used to automatically generate the taxonomies of the nominal attribute values for a real world database. An ontology for a specific analysis task is finally constructed, which reflects some interesting behaviour of real data. Although this semi-automatically constructed ontology may be different from the widely accepted one for the same domain, it may indicate the true character of the data from the statistical point of view and have a semantic interpretation as well as being more suitable for the specific data mining application.