Geographically Weighted Machine Learning Model for Addressing Spatial Heterogeneity of Public Health Development Index in Java Island

Aug 5, 2026

Upload by Oliver

Random Forest (RF) improves prediction but uses a global tree structure that may miss geographic variations when modeling public health. Relationships between the Public Health Development Index (PHDI) and district risk factors can differ across space. This study applies a modified Geographically Weighted Random Forest (GW-RF) to calibrate locally and compare its results with a global RF baseline for Java Island. Outcomes assess PHDI using six district-level indicators and reveal strong ability to explain spatial heterogeneity and guide geographic targeting.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Introduction

Spatial heterogeneity and geographic context

Limits of conventional ML and global RF

Geographically weighted regression (GWR) and its drawbacks

Geographically weighted random forest (GW-RF) approach

GEOGRAPHICALLY WEIGHTED MACHINE LEARNING MODEL FOR ADDRESSING SPATIAL HETEROGENEITY OF PUBLIC HEALTH DEVELOPMENT INDEX IN JAVA ISLAND Muhammad Azis Suprayogi 1*, Bagus Sartono2, Khairil Anwar Notodiputro3 1,2,3Department of Statistics, Faculty of Mathematics and Natural Science, IPB University Jalan Meranti, Kampus IPB Dramaga, Bogor, West Java, 16680, Indonesia Corresponding author’s e-mail: * [azissuprayogi@apps.ipb.ac.id](azissuprayogi@apps.ipb.ac.id) ABSTRACT Article History: Received: 16th, May 2024 Revised: 4th, July 2024 Accepted: 9th,August 2024 Published:14th, October 2024 Keywords: Geographically Weighted; GW-RF; Random Forest; Spatial. Random Forest (RF) machine learning models have emerged as a prominent algorithm, addressing problems arising from the sole use of decision trees, such as overfitting and instability. However, conventional RF has global coverage that may need to capture spatial variations better. Based on the analysis of the level of public health development, the relationship between the level of health development and risk factors can vary spatially. We use a modified RF algorithm called Geographically Weighted Random Forest (GW-RF) to address this challenge. GW-RF, as a tree-based non-parametric machine learning model, can help explore and visualize relationships between the Public Health Development Index (PHDI) as response variables and factors that are indicatorsat the district level. GW-RF output is compared with global output, which is RF in 2018 using the percentage of the population with access to clean/decent water (X1), consumption of eggs and milk per capita per week (X2), number of healthcare facilities per 1000 people (X3), number of doctors per 1000 people (X4), pure participation rate ratio female/male (X5), percentage of households that have hand washing facilities with soap and water (X6) as independent variables. Our results show that the non-parametric GW-RF model shows high potential for explaining spatial heterogeneity and predicting PHDI versus a global model when including six major risk factors. However, some of these predictions mean little. Findings of spatial heterogeneity using GW-RF show the need to consider local factors in approaches to increasing PHDI values. Spatial analysis of PHDI provides valuable information for determining geographic targets for areas whose PHDI values need to be improved. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution-ShareAlike 4.0 International License. How to cite this article: M. A. Suprayogi, B. Sartono, and K. A. Notodiputro.,“GEOGRAPHICALLY WEIGHTED MACHINE LEARNING MODEL FOR ADDRESSING SPATIAL HETEROGENEITY OF PUBLIC HEALTH DEVELOPMENT INDEX IN JAVA ISLAND,” BAREKENG: J. Math. & App., vol. 18, iss. 4, pp. 2577-2588, December, 2024. Copyright © 2024 Author(s) Journal homepage: [https://ojs3.unpatti.ac.id/index.php/barekeng/](https://ojs3.unpatti.ac.id/index.php/barekeng/) Journal e-mail: [barekeng.math@yahoo.com](barekeng.math@yahoo.com); [barekeng.journal@mail.unpatti.ac.id](barekeng.journal@mail.unpatti.ac.id) Research Article ∙ Open Access 1. INTRODUCTION Machine Learning (ML) models have high predictive capabilities from data mining and are often flexible and non-linear. However, it is often less than optimal in capturing geographic relationships, making it less sensitive to spatial context. This challenge is significant given that spatial data usually exhibit heterogeneity, leading to variations in the relationship between dependent and independent variables across regions. Conventional ML models need help dealing with such complexity because they produce a single output for the entire study area without considering the spatial variations that may exist. Research on handling spatial heterogeneity in population modeling based on geographic data is still limited [1] . Previous research on introducing a framework for mod

FAQs

Why can a conventional global Random Forest be insufficient for public health spatial data?

It produces a single model for the entire area, so it may not capture how relationships between PHDI and risk factors vary across regions.

What is the key idea behind Geographically Weighted Random Forest (GW-RF)?

GW-RF calibrates models locally rather than globally, decomposing a global process into spatially varying sub-models to better reflect geographic heterogeneity.

Which variables are used to predict the Public Health Development Index in the comparison?

The model uses six district-level risk factors: access to clean/decent water (X1), eggs and milk consumption per capita per week (X2), healthcare facilities per 1000 people (X3), doctors per 1000 people (X4), female/male participation rate ratio (X5), and households with hand-washing facilities with soap and water (X6).

Aug 5, 2026

Upload by Oliver