On fuzzy clustering based regression models
Mika Sato‐Ilic · IEEE Annual Meeting of the Fuzzy Information, 2004. Processing NAFIPS '04. · 2004
We have proposed a fuzzy cluster loading [7] which can show the relationship between the obtained fuzzy clusters and the variables in order to interpret the obtained fuzzy clusters. Moreover, we have proposed a weighted regression model using a fuzzy clustering result obtained by the classification of the data with respect to explanatory variables. [8] These models are closely related with the conventional geographically weighted regression model [2] and the fuzzy c-regression model. [5] So, this paper discusses the difference and the relation of these models from the view point of the difference of the estimates of the regression coefficients and the assumption of the errors. Several numerical examples show the difference and the better performance of the proposed models.