Fuzzy modeling and similarity analysis applied to ecological data

Fuzzy rule-based models offer a suitable framework for the modeling of ecological systems The complexity of such systems requires considerable idealization for mechanistic modeling to be applicable. This reduces the numeric accuracy while the interpretability of the models remains difficult In this paper a Takagi-Sugeno (TS) fuzzy model with linear consequents is used to model the algae growth in lakes. Both the membership functions in the premise and the consequent parameters are estimated from measurements of relevant quantities by means of product-space fuzzy clustering. To enhance the interpretability of the model, similarity analysis is applied and similar fuzzy sets and rules are combined, giving a transparent and compact model without notably altering the accuracy.