Assessing the Predictive Power of Urban Green Spaces in Machine Learning Models for Chicago Housing Prices

Aug 3, 2026

Upload by Sage

Urban Green Space (UGS) is examined as a key urban environmental factor influencing housing values. The study compares the predictive power of two UGS measurement approaches—park proximity and NDVI—in machine learning models for Chicago housing prices. Using real estate big data, GIS analysis, and machine learning, the findings indicate that NDVI serves as a stronger predictor for single-family residential properties located near the urban core fringe.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages29

Reading Time1 h 13 min

AccessFree to download

Abstract

1 Introduction

2 Literature Review

THE UNIVERSITY OF CHICAGO Assessing the Predictive Power of Urban Green Spaces in Machine Learning Models for Chicago Housing Prices By Kuang Sheng May 2025 A paper submitted in partial fulfillment of the requirements for the Master of Arts degree in the Master of Arts in Computational Social Science Faculty Advisor: Crystal Bae Preceptor: Fabricio Vasselai Abstract Urban Green Space (UGS) plays an important role in the urban environment. This study compares the predictive power of two types of UGS measurements—park proximity and NDVI — in the machine learning models to predict housing prices in Chicago. By incorporating real estate big data, GIS analysis, and Machine Learning techniques, the result indicates NDVI is a strong predictor for single-family residential properties on the fringe of the urban core. Keywords: Computational Social Science; Urban Geography; GIS; Machine Learning; Real Estate 1 Introduction Although Urban Green Space (UGS) is a fundamental part of the urban environment that generates economic benefits (Kim & Peiser, 2018), supports urban residents’ physical and mental health (Russo & Cirella, 2018), and serves as a city landmark that constitutes the city image (Groos & Dages, 2008), the definition of UGS still shows significant inconsistency across different disciplines (Taylor & Hochuli, 2017) . There are many great examples of centrally planned UGS implementation in global cities: Central Park in Manhattan, NY; Millennium Park in Chicago, IL; Discovery Green in Houston, TX; Ueno Park in Tokyo, JP; and Hyde Park in London, UK. At the same time, accessible UGS at a lower hierarchical level also contributes to the urban ecosystem (Gupta et al. , 2016) . In academia, UGS has received significant attention from scholars in recent years due to the COVID-19 pandemic, bringing greater public awareness to the value of both physical and mental health. Researchers conducted experiments in different locations around the world to evaluate if and how interacting with UGS can alleviate stress and enhance mental wellness during the pandemic lockdown. However, most research projects solely rely on a single source of green space data: government data portals that include only governmentregistered park space, which are comparatively less comprehensive due to the exclusion of minor and private green spaces, like sidewalk trees, grassy medians, and private backyards, potentially leading to data deficiency and inaccuracy. Such a data collection methodology might underestimate the value of UGS in machine learning models, which have been widely used to predict housing prices in the real estate market. At the same time, there is another type of UGS data—the Normalized Difference Vegetation Index (NDVI) . NDVI is a remote sensing index that measures vegetation health and density based on remote sensing analysis of satellite imagery, providing more compre- hensive and geographically continuous information about UGS distribution and quality. Intoday’s real estate industry, platforms such as Zillow and Redfin have integrated machine learning-based price estimations as a key feature to enhance property valuation and user decision-making. Accurate price estimation contributes to a more transparent housing market, enhances the overall home-buying experience, and promotes social equity by reducing asymmetric information and supporting equal access to housing opportunities. I hypothesize that compared to government data on park proximity, NDVI provides better predictive power in machine learning models. To this end, I pose the research question: How do the two different methods of measuring UGS—distance to park space and ND VI—influence the predictive performance of machine learning models in estimating Chicago housing prices? For this graduate thesis, I first demonstrated the importance and significance of UGS research in social science disciplines, including geography, psychology, and urban studies

FAQs

What are the two ways the study measures Urban Green Space (UGS)?

The study compares park proximity and NDVI as two different UGS measurement methods for predicting housing prices in Chicago.

Which UG S measurement shows stronger predictive power in the results?

NDVI is reported as a strong predictor, particularly for single-family residential properties on the fringe of the urban core.

What data and methods are used to evaluate housing price prediction?

The research uses Chicago listing data from Redfin, computes housing intrinsic features and environmental features via GIS, and applies mainstream machine learning models with explainable machine learning techniques for comparison.

Aug 3, 2026

Upload by Sage