Ontology-Driven Knowledge Discovery

The problem of Knowledge Discovery has always attractedmany researchers and continues to be of great relevance to thecomputer science community in the branch of learning. Thisthesis aims to contribute to this topic, getting hints from theOntologyandData Miningenvironments.We investigate a method for extracting new implicit knowl-edge directly from an ontology by using an inductive/deductiveapproach. By giving a sort ofBayesianinterpretation to rela-tionships that already exist in an ontology, we are able to re-turn the extracted knowledge in form ofInfluence Rules.The idea is to split the extraction process in two separate phasesby exploiting the ontology peculiarity of keeping metadata(the schema) and data (the instances) separate. The deduc-tive process draws inference from the ontology structure, bothconcepts and properties, by applying link analysis techniquesand producing a sort of implications (rules schemas) in whichonly the most important concepts are involved. Then an in-ductive process, realized by a data mining algorithm, exploresthe ontology instances for enriching the implications and build-ing the final rules. What we want to prove, besides the correctness and feasibil-ity1of the project, is that the approach allows us to extract"higher level" rules w.r.t. classical knowledge discovery tech-niques. In fact, ontology metadata gives a general view of thedomain of interest and supplies information about all the ele-ments apart from the fact that they are included as instancesin the collected data. The technique is completely generaland applicable to each domain. Since the output is a set of"standard" Influence Rules, it can be used to integrate existingknowledge or for supporting any other data mining process.