Incremental membership function updates

Abstract

Many fuzzy applications today are based on large databases that change dynamically. Particularly, in many flexible querying systems this represents a huge problem, since changing data may lead to poor results in the absence of proper retraining. In this paper we propose a novel incremental approach to represent the membership functions describing the linguistic terms for a dynamically changing database. It exploits fuzzy knowledge models previously determined in order to simplify the modelling process. Experiments testing the method's efficiency are also reported. © Springer-Verlag Berlin Heidelberg 2010.