A Computatioal Analysis of Kernel-Based Nonparametric Regression Applied to Poverty Data

Abstract

This research aims to model the relationship between poverty and socioeconomic variables in Nusa Tenggara Timur Province, Indonesia. The purpose of the study is to assess the effectiveness of nonparametric regression, specifically using kernel methods, to provide a more accurate representation of the complex and nonlinear relationships between predictor variables and poverty levels. The study focuses on several key variables, including average years of schooling, labor force participation rate, percentage of households with access to electricity, population density, illiteracy rate, and life expectancy. The research utilized a kernel regression approach, comparing the performance of different kernel functions, including Gaussian, Epanechnikov, Triangle, and Quartic kernels. The model’s performance was evaluated using metrics such as Mean Squared Error (MSE), Generalized Cross Validation (GCV), and the coefficient of determination (R²). The results showed that the Gaussian kernel function provided the most accurate predictions for poverty levels, with the best balance between model complexity and error.