Pattern-based clustering for database attribute values

摘要: We present a method for automatically clustering similar attribute values in database system spanning mulitple domains. The constructs an abstraction hierarchy each using rules that are derived from the instance. have confidence and popularity combine to express "usefullness" of rule. Attribute clustered if they used as premise with same consequence. By iteratively applying algorithm, clusters can be found. algorithm improved by allowing domain expert supervision during process. An example well experimental results large transportation included.