Analysis of Ozone Concentration in Wasit Governorate – Iraq Using Fuzzy Nonparametric Regression Estimation Methods

Abstract

    This article aims to analyze the relationship between temperature and ozone concentration in Wasit Governorate using fuzzy nonparametric regression models, given the nonlinear and ambiguous nature of the environmental data. Three main fuzzy nonparametric regression modeling methods were employed: the Nearest Neighbor (KNN) method, the Nadaraya–Watson method, and the Local Polynomial method. The estimation process relied on the Leave-One-Out (LOOCV) cross-validation method to select the optimal parameters for each model, using the Mean Squared Error (AMSE) and the Mean Integral Squared Error (IMSE) criteria. The response variable was treated as a triangular fuzzy number representing the uncertainty in the environmental measurements, and the Diamond squared distance was used to measure the difference between the estimated and actual fuzzy numbers. The results showed that all models performed acceptably, but the local linear regression model with the Epanechnikov kernel outperformed by achieving the lowest AMSE and IMSE values, reflecting its higher accuracy in representing the nonparametric relationship between temperature and ozone concentration. These results confirm that nonparametric fuzzy models represent environmental phenomena characterized by ambiguity.