KDD Support Services Based on Data Semantics

The identification of valid, novel and interesting models from large volumes of data is the primary goal of Knowledge Discovery in Databases (KDD). In order to successfully achieve such a complex goal, many kinds of semantic information about the KDD and business domains is necessary. In this paper, we present an approach to the characterization of semantic domain information for a particular kind of KDD process: classification. In particular we show how, by estimating the properties of the true but unknown classification model, one can derive domain information on the classification problem at hand. We discuss how, by saving these properties with the data, users profit from this information and save time for experimenting with a lot of classifiers and parameters by accessing this knowledge.