Econometric Modelling based on Pattern recognition via the Fuzzy c-Means Clustering Algorithm

In this paper we consider the use of fuzzy modelling in the context of econometric analysis of both time-series and cross-section data. We discuss and demonstrate a semi-parametric methodology for model identification and estimation that is based on the Fuzzy c-Means algorithm that is widely used in the context of pattern recognition, and the Takagi-Sugeno approach to modelling fuzzy systems. This methodology is exceptionally flexible and provides a computationally tractable method of dealing with non-linear models in high dimensions. In this respect it has distinct theoretical advantages over non-parametric kernel regression, and we find that these advantages also hold empirically in terms of goodness-of-fit in a selection of economic applications.