# RECORD 2cdb6f374417590715ab5db24f593d17616b9ea2
# TITLE: Effects of neighborhood streetscape on the single-family housing price: Focusing on nonlinear and interaction effects using interpretable machine learning
# YEAR: 2025.0  VENUE: PLoS ONE  DOI: 10.1371/journal.pone.0323495
# AUTHORS: J. Han; Ayoung Woo; Sugie Lee
# PAGES: 25  CANDIDATE TABLE PAGES (1-based): [13]
# ABSTRACT: Previous studies using the conventional Hedonic Price Model to predict existing housing prices may have limitations in addressing the relationship between house prices and streetscapes as visually perceived at the human eye level, due to the constraints of streetscape estimations. Therefore, in this study, we analyzed the relationship between streetscapes visually perceived at eye level and single-family home prices in Seoul, Korea, using computer vision technology and machine learning algorithms. We used transaction data for 13,776 single-family housing sales between 2017 and 2019. To measure visually perceived streetscapes, this study used the Deeplab V3 + deep-learning model with 233,106 Google Street View panoramic images. Then, the best machine-learning model was selected by comparing the explanatory powers of the hedonic price model and all alternative machine-learning models. According to the results, the Gradient Boost model, a representative ensemble machine learning model, performed better than XGBoost, Random Forest, and Linear Regression models in predicting single-family house prices. In addition, this study used an interpretable machine learning model of the SHAP method to identify key features that affect single-family home price prediction. This solves the “black box” problem of machine learning models. Finally, by analyzing the nonlinear relationship and interaction effects between perceived streetscape characteristics and house prices, we easily and quickly identified the relationship between variables the hedonic price model partially considers.


===== PAGE 1 (score 6, numeric-paren 0, table-words 0, var-words 6) =====

PLOS One | https://doi.org/10.1371/journal.pone.0323495  May 21, 2025
1 / 25
 
 OPEN ACCESS
Citation: Han J, Woo A, Lee S (2025) Effects of 
neighborhood streetscape on the single-family 
housing price: Focusing on nonlinear and 
interaction effects using interpretable machine 
learning. PLoS One 20(5): e0323495. https://
doi.org/10.1371/journal.pone.0323495
Editor: Zeyar Aung, Khalifa University, UNITED 
ARAB EMIRATES
Received: June 18, 2024
Accepted: April 8, 2025
Published: May 21, 2025
Copyright: © 2025 Han et al. This is an open 
access article distributed under the terms of 
the Creative Commons Attribution License, 
which permits unrestricted use, distribution, 
and reproduction in any medium, provided the 
original author and source are credited.
Data availability statement: Please note 
that our data is available at the following 
link (Figshare) https://doi.org/10.6084/
m9.figshare.26965252.
Funding: This work was supported by the 
National Research Foundation of Korea (NRF) 
grant funded by the Korea government (*MSIT) 
RESEARCH ARTICLE
Effects of neighborhood streetscape on 
the single-family housing price: Focusing 
on nonlinear and interaction effects using 
interpretable machine learning
Jaewon Han1, Ayoung Woo
2, Sugie Lee
1*
1  Department of Urban Planning & Engineering, Hanyang University, Seoul, South Korea, 2  Graduate 
School of Urban Studies, Hanyang University, Seoul, South Korea 
* sugielee@hanyang.ac.kr
Abstract 
Previous studies using the conventional Hedonic Price Model to predict existing 
housing prices may have limitations in addressing the relationship between house 
prices and streetscapes as visually perceived at the human eye level, due to the 
constraints of streetscape estimations. Therefore, in this study, we analyzed the 
relationship between streetscapes visually perceived at eye level and single-family 
home prices in Seoul, Korea, using computer vision technology and machine learning 
algorithms. We used transaction data for 13,776 single-family housing sales between 
2017 and 2019. To measure visually perceived streetscapes, this study used the 
Deeplab V3 + deep-learning model with 233,106 Google Street View panoramic 
images. Then, the best machine-learning model was selected by comparing the 
explanatory powers of the hedonic price model and all alternative machine-learning 
models. According to the results, the Gradient Boost model, a representative ensem­
ble machine learning model, performed better than XGBoost, Random Forest, and 
Linear Regression models in predicting single-family house prices. In addition, this 
study used an interpretable machine learning model of the SHAP method to identify 
key features that affect single-family home price prediction. This solves the “black 
box” problem of machine learning models. Finally, by analyzing the nonlinear relation­
ship and interaction effects between perceived streetscape characteristics and house 
prices, we easily and quickly identified the relationship between variables the hedonic 
price model partially considers.
1.  Introduction
The estimation of housing prices has been a longstanding topic for urban planners 
and policymakers, as housing units are one of the most expensive commodities in 
household finances. Housing prices are closely related to macroeconomic conditions, 


===== PAGE 13 (score 32, numeric-paren 0, table-words 4, var-words 17) =====

PLOS One | https://doi.org/10.1371/journal.pone.0323495  May 21, 2025
13 / 25
Second, latitude (north-south) and longitude (east-west), which represent the location properties of the houses, had 
negative and positive effects on the predicted price of single-family houses, respectively. The polarization of the housing 
supply within Seoul and the differences in urban infrastructure are severe, and those differences are reflected in housing 
prices. In addition, in the local Shapley value graph, the red and blue dots are mixed, which indicates that a non-linear 
relationship exists between location (latitude and longitude) and single-family house prices.
Table 1.  Definitions of the variables and basic statistics.
Variables
Measurements
Mean
S.D.
Min.
Max.
Dependent variable
  ln (housing prices)
The natural logarithm of the price of a single-family house (KRW)
20.17
0.55
17.32
21.24
Independent variables
Structural features
  Site area
The site area of a single-family house (m2)
140.33
1,875.98
3.00
220,220
  Gross floor area
The total floor area of a single-family house (m2)
122.63
70.21
4.50
479.94
  Housing age
Age of the building
39.70
11.06
0
95
  Building structure:
Ferroconcrete
Ferroconcrete = 1, others = 0
0.01
0.12
0
1
  Building structure:
Masonry
Masonry = 1, others = 0
1.46e-03
0.04
0
1
  Land topography
Flatland = 1, others = 0
0.67
0.47
0
1
  Site shape
Regular form = 1, others = 0
0.84
0.37
0
1
  Corner parcel
Corner parcel = 1, others = 0
0.13
0.33
0
1
  Parking availability
On-site parking allowed = 1, on-site parking not allowed = 0
0.01
0.12
0
1
Walking environment
Macro-scale walking environment factors
  3-way intersection
density
(Number of 3-way intersections/total street length) within the 
400m buffer
3.29E-03
1.59e-03
0
0.02
  4-or-more-way
intersection density
(Number of 4-or-more-way intersections/total street length) within 
the buffer
6.85e-04
5.34e-04
0
3.51e-03
  Subway station density
(Number of subway stations/total street length) within the buffer
5.73e-04
9.55e-04
0
0.01
  Bus stop density
(Number of bus stops/total street length) within the buffer
0.01
0.05
0
3.56
  Crosswalk density
(Number of crosswalks/total street length) within the buffer
0.02
0.06
0
7.34
  Park land use
Total park and open space area within the buffer (m2)
44,431.24
66,091.75
0
447,553.80
  Residential land use
Total area for residential use within the buffer (m2)
441,246.40
183,936.50
1,615.14
1,421,861.00
  Business land use
Total area for businesses within the buffer (m2)
19,325.62
40,873.86
0
917,400.30
  Commercial land use
Total area for commercial use within the buffer (m2)
73,537.93
54,375.07
0
681,313.20
  Mixed land use
–
k
∑
i=1
(pi) ∗ln(pi)
ln(k)
pi=Total area of land use in the 400m buffer
0.15
0.02
0.97
0.97
k=Number of land uses
Streetscape environment
  ln (greenery)
The natural logarithm of the average greenery value within the buffer
1.63
0.46
0.18
3.93
  ln (enclosure)
The natural logarithm of the average enclosure value within the 
buffer
4.69
0.22
1.93
5.14
  ln (paved sidewalks)
The natural logarithm of the average paved sidewalk value within 
the buffer
2.52
0.12
2.01
3.13
Locational characteristics
  ln (distance to parks)
The natural logarithm of the distance to the nearest park 
boundary (m)
4.86
1.00
0
6.95
  Distance to CBD
The shortest distance to the CBD boundary (m)
3,765.00
2,242.81
0
11,539.12
https://doi.org/10.1371/journal.pone.0323495.t001


===== PAGE 14 (score 18, numeric-paren 0, table-words 2, var-words 13) =====

PLOS One | https://doi.org/10.1371/journal.pone.0323495  May 21, 2025
14 / 25
Third, the distance from the city center and the green area around each single-family house, which also represent 
location characteristics, had a negative effect on the predicted price of single-family houses. The CBD and parks are 
perceived negatively in a residential environment because they draw many people from outside the neighborhood. On the 
other hand, the business district area, commercial district area, and subway station density had positive effects because 
the various benefits provided by the commercial and business facilities that concentrate around subway stations. In the 
local Shapley value graph, the red and blue dots are mixed, indicating a non-linear relationship between the location and 
environment around a neighborhood and single-family house prices within it.
Fourth, bus stops and land use mix, which are closely related to walkability, had negative and positive effects on hous­
ing prices, respectively. Many people visit a bus stop, which are often near busy streets and experience various pollutants 
Table 2.  Performance comparison of machine learning models.
Model
Performance evaluation indicator
R2
MSE
MAE
MedAE
Linear Regression
0.660
0.104
0.248
0.197
Random Forest
0.318
0.209
0.363
0.318
XGBoost
0.765
0.072
0.200
0.155
Gradient Boost
0.817
0.056
0.169
0.122
https://doi.org/10.1371/journal.pone.0323495.t002
Fig 6.  Analysis of variable importance to single-family housing prices using the SHAP method.
https://doi.org/10.1371/journal.pone.0323495.g006