An XML-Based Database for Knowledge Discovery Rosa Meo 1 and Giuseppe Psaila 2 1 Universit`a di Torino, Dipartimento di Informatica, corso Svizzera 185, I-10149, Torino, Italy meo@di.unito.it 2 Universit`a di Bergamo, Facolt`a di Ingegneria, Viale Marconi 5, I-24044 Dalmine (BG), Italy psaila@unibg.it Abstract. Pattern Management Systems and Inductive Databases, are proposed as a new generation of general purpose databases with the aim to manage data mining patterns and work as knowledge bases in sup- port to the deployment of the KDD process. One of the main problems to be solved is the integration between data and patterns and pattern maintenance when data update. Unfortunately, the heterogeneity of the patterns that represent the extracted knowledge and of the different con- ceptual tools used to find the patterns make difficult this integration in a unique framework. In this paper, we explore the feasibility of using XML as the unifying framework for inductive databases, and present a model, named XDM (XML for Data Mining). We will show the basic features of the model, such as the storage in the same database of both data and patterns. To store patterns, we consider determinant for their interpretation the stor- age of the pattern derivation process which is described by the concept of statement, based on data mining operators. Some of the statements are automatically generated by the system while maintaining consistence between source and derived data. Furthermore, we show how the use of XML namespaces allows the effective coexistence of different data mining operators and provides extensibility to new operators. Finally, we show that with the use of XML-Schema we are able to define the schema, the state and the integrity constraints of an inductive database. 1 Introduction Data mining applications are called to extract descriptive and predictive pat- terns, typically used for decision making, from the data contained in traditional databases and from other unconventional information systems such as the web. Inductive Databases (IDB) have been launched in [8] as general-purpose data- bases in which both the data and the patterns can be represented, retrieved and manipulated with the goal to assist the deployment of the Knowledge Dis- covery Process (KDD). Thus, KDD becomes a querying sequence in a query T. Grust et al. (Eds.): EDBT 2006 Workshops, LNCS 4254, pp. 814–828, 2006. c  Springer-Verlag Berlin Heidelberg 2006