Clustering-based identification of TS-models: comparison on a groundwater model case study

We apply different clustering algorithms for the identification of Takagi-Sugeno models. All of the fuzzy c-means, Gustafson-Kessel, simplified Gustafson-Kessel, Gath and Geva, simplified Gath and Geva, and modified Gath and Geva clustering algorithms try to minimize the same objective function. First, an algorithm for determining the optimal number of clusters is presented. The Takagi-Sugeno models with the optimal number of clusters are then incorporated into a groundwater model, and compared with measurements of the EMSL experiment and the results of a numerical groundwater model.