COMPARATIVE ANALYSIS FOR FUZZY NONPARAMETRIC REGRESSION MODELS

Statistical modeling is essential to revealing the relationships between variables. These statistical models can be classified as parametric and nonparametric methods in studies using crisp values. However, most of the collected data are inherently fuzzy. In this context, the fuzzy expression of methods using precise data is a matter of curiosity for researchers. The methods with fuzzy input and output variables have been developed for a long time. The study aims to describe nonparametric local polynomial regression models in fuzzy structure to examine the results for cases where the input variable is a crisp number, and the output variable is a symmetrical triangular and trapezoidal fuzzy number. According to the results, the bandwidth parameter was smaller in models where the degree of the polynomial was taken as one and larger in the case of three. In addition, the bandwidth parameter was found to be larger in models using the Epanechnikov kernel.