Zusammenfassung
this paper, we will describe a methodology and a set
of computerized tools that discover characteristic
queries as discriminant rules from structured
databases. In order to discover useful set of
discriminant characteristic queries from a database,
our approach is first to cluster the target database
and to discover a set of queries from each cluster.
Figure 1 depicts the overall steps in our methodology.
First, the input data set that user is interested in is
selected, preprocessed, and represented in a proper
format for clustering process. Second, generalized
clustering algorithms are then applied to the
preprocessed data set grouping the target database into
clusters of objects with similar properties. In the
third step, we try to characterize the clusters found
in the clustering process. For this purpose, we use a
query discovery system called MASSON that uses database
queries as its rule representation language 9. MASSON
discovers a discriminant query (or set of queries) that
describes a given set of objects in databases using
genetic programming (GP) 4. The discovered query or a
set of queries can distinctively describe the
commonalities for the given set of objects with respect
to the other objects in a database 8,10.
Nutzer