Using Prior Knowledge in Data Mining

One of the most important and challenging problems in current Data Mining research is the definition of the prior knowledge that can be originated from the process or the domain. This contextual information may help select the appropriate information, features or techniques, decrease the space of hypotheses, represent the output in a most comprehensible way and improve the process. Ontological foundation is a precondition for efficient automated usage of such information (Chandrasekaran et al., 1999). An ontology is a formal explicit specification of a shared conceptualization for a domain of interest (Gruber, 1993). Among other things, this definition emphasizes the fact that an ontology has to be specified in a language that comes with a formal semantics. Due to this formalization ontologies provide the machine interpretable meaning of concepts and relations that is expected when using a semantic-based approach (Staab & Studer, 2004). In its most prevalent use in Artificial Intelligence (AI), an ontology refers to an engineering artifact (more precisely, produced according to the principles of Ontological Engineering (Gómez-Pérez et al., 2004)), constituted by a specific vocabulary used to describe a certain reality, plus a set of explicit assumptions regarding the intended meaning of the vocabulary words. This set of assumptions has usually the form of a First-Order Logic (FOL) theory, where vocabulary words appear as unary or binary predicate names, respectively called concepts and relations. In the simplest case, an ontology describes a hierarchy of concepts related by subsumption relationships; in more sophisticated cases, suitable axioms are added in order to express other relationships between concepts and to constrain their intended interpretation. Ontologies can play several roles in Data Mining (Nigro et al., 2007). In this chapter we investigate the use of ontologies as prior knowledge in Data Mining. As an illustrative case throughout the chapter, we choose the task of Frequent Pattern Discovery, it being the most representative product of the cross-fertilization among Databases, Machine Learning and Statistics that has given rise to Data Mining. Indeed it is central to an entire class of descriptive tasks in Data Mining among which Association Rule Mining (Agrawal et al., 1993; Agrawal & Srikant, 1994) is the most popular. A pattern is considered as an intensional description (expressed in a given language L) of a subset of a data set r. The support of a pattern is the relative frequency of the pattern within r and is computed with the evaluation function supp. The task of Frequent Pattern Discovery aims at the extraction of all frequent patterns, i.e. all patterns whose support exceeds a user-defined threshold of minimum support. The blueprint of most algorithms for Frequent Pattern Discovery is the levelwise search (Mannila & Toivonen, 1997). It is based on the following assumption: If a generality order = for the language L of patterns can be found such that = is monotonic w.r.t. supp, then the resulting space (L, =) can be searched breadth-first by starting from the most general pattern in L and alternating candidate generation and candidate evaluation phases.