Additive kernel-based nonlinear regression for fuzzy panel data

Gholamreza Hesamian

doi:10.1109/tfuzz.2026.3716281

Additive kernel-based nonlinear regression for fuzzy panel data

2026-07-22

0

PRE

AI

DOI：10.1109/tfuzz.2026.3716281

Original

Original request for help

Save

Abstract

En 中文

Panel regression is a widely used statistical frame work for analyzing data that combine cross-sectional and time series dimensions. In many real-world applications, however, the response variables are subject to vagueness, measurement uncertainty or linguistic assessment, making common crisp panel models inadequate. Existing fuzzy panel regression approaches have primarily focused on linear specifications, which may be overly restrictive when the underlying relationships are complex and inherently nonlinear. To address these limitations, this paper introduces a flexible fuzzy nonlinear panel regression model that captures nonlinear associations between fuzzy responses of general LR-type and crisp predictors. The proposed methodology is based on an additive nonparametric kernel framework, where the unknown smooth component functions are estimated using the Nadaraya-Watson estimator in a two-step procedure. This approach extends existing fuzzy linear panel models by incorpo rating a nonlinear structure while maintaining computational tractability and broad applicability. The effectiveness of the proposed model is demonstrated through a simulation study and three real-world panel data applications containing comparative analysis with established fuzzy linear panel regression methods. The empirical results show that the proposed fuzzy nonlinear model achieves superior fitting and predictive accuracy.

Keywords:

Fuzzy panel data

nonlinear model

LR fuzzy number

nonparametric estimator

kernel

performance measure

Additive kernel-based nonlinear regression for fuzzy panel data

Abstract

Journal

Organization

Cited Papers