Gaussian Process to Takagi-Sugeno Fuzzy Model Using Supervised Clustering

In this paper, we present a novel approach for transforming Gaussian process models into Takagi-Sugeno fuzzy models (GP2TS). Our approach utilizes a supervised clustering algorithm to gradually merge data points into larger and larger clusters based on the errors of the local models and the prediction intervals that capture the underlying nonlinearities in the data. This results in a reduced number of local models that accurately describe the problem in their respective vicinity. Compared to the Gaussian process models, our GP2TS approach offers several advantages. First, it enhances the interpretability of local models, making it easier to understand the behavior of the model and its predictions. Secondly, it improves the applicability of the model by reducing computational demand, making it more practical for real-world applications. To demonstrate the potential of our proposed approach, we apply it to two benchmark problems from the literature. The results show that the GP2TS approach performs well in practice and can achieve similar levels of accuracy compared to other state-of-the-art fuzzy model identification techniques.