Partition identification of fuzzy models using objective function clustering algorithms

The identification of the partitioning of the input space is a difficult but important aspect concerning the identification of fuzzy models. This article discusses how to apply objective function cluster algorithms such as the fuzzy-c-means for this task. Fuzzy models with multidimensional reference fuzzy sets are considered which provide for good model performance and enable an automated identification procedure. Some guidelines are presented and the choice of the parameters of the clustering algorithms is discussed. The goal of the article is not to present the results for a particular system but to give structured advice to those who want to identify fuzzy models with high accuracy for their applications