Eyke Hüllermeier
- “Fuzzy sets in association analysis (Data Mining): Useful or Not?”
Abstract
In order to allow for the analysis of data sets including numerical attributes, several generalizations of association rule mining based on fuzzy sets have been proposed in literature. There are different motivations for a fuzzy approach to association analysis, notably the following: Firstly, by allowing for “soft” rather than crisp boundaries of interval, fuzzy sets can avoid certain undesirable threshold or boundary effects. Secondly, fuzzy association rules are very appealing from a knowledge representational point of view: The very idea of fuzzy sets is to act as an interface between a numerical scale and a symbolic scale which is usually composed of linguistic terms. Thus, the rules discovered in a database might be presented in a linguistic and hence comprehensible and user-friendly way. Thirdly, the possibility of mining different types of fuzzy rules with different semantics makes association analysis more expressive. On the other hand, these potential advantages do not come for free and, perhaps even more importantly, have not been proved convincingly from a practical point of view. It is hence hardly astonishing that the usefulness of fuzzy association analysis has recently been called into question by some researchers. In this talk, I will discuss advantages and disadvantages of fuzzy association analysis in general and respond to the aforementioned criticisms in particular.
University of Magdeburg (Germany)
Department of Computer Science
Universitâtsplatz 2, 39106 Magdeburg, Germany
huellerm@iti.cs.uni-magdeburg.de