Information : cette page n'est pas traduite en français
2017-03-24 An Exploratory Knowledge Discovery Process based on Formal Concept Analysis
An Exploratory Knowledge Discovery Process based on Formal Concept Analysis, Seminar room, LIRMM building 4
Amedeo Napoli (DR CNRS - LORIA, Nancy)
Abstract: Knowledge discovery (KD) in complex datasets can be considered from several points of view, e.g. data, knowledge, and problem solving. KD is applied to datasets and has a direct impact on the design of knowledge bases and problem solving. Dually, principles supporting knowledge representation can be reused in knowledge discovery, as in declarative data mining, for making KD more exploratory, iterative and interactive. Accordingly, one of our main objectives in KD is to design an exploratory KD approach supporting interactions between data and domain knowledge, discovery and representation, and between the analyst and the KD process. Accordingly, in this presentation, we discuss the process of Exploratory Knowledge Discovery based on Formal Concept Analysis (FCA). FCA starts with a binary context and outputs a concept lattice, which can be visualized, navigated and interpreted by human agents, and, as well, which can be processable by software agents. FCA can be extended to pattern structures for dealing with complex data such as Linked Data (RDF). We will present a methodology based on FCA and pattern structures for mining definitions in RDF data. Such data are classified into a concept lattice allowing data exploration and the discovery of implications, used to automatically detect missing information and to complete RDF data. Contrasting the preceding data-directed KD process, we will present a goal-directed KD process, where the search for patterns of high interest is guided by the analyst preferences and domain constraints. Finally, we will conclude by discussing the benefits of FCA and pattern structures for exploratory knowledge discovery.
Dernière mise à jour le 10/03/2017