Research News

A hedonic price model incorporating street view images processed by machine learning and existing property data achieves nearly 75% accuracy for rent prediction in Osaka City

Factors in rent prices

A new method predicts rents with high accuracy by adding variables of streetscape components and neighborhood perceptions to an existing hedonic price model.

Credit: Osaka Metropolitan University

Housing rents usually correlate with factors such as the building’s age, facilities, and location. Yet not all rentals with similar physical factors charge the same rent. Psychological factors such as the subjective perceptions of the neighborhood matter as well.

Considering these perception variables, an Osaka Metropolitan University team has developed a method with almost 75% accuracy in explaining housing prices in Osaka City.

The team led by Graduate School of Human Life and Ecology student Xiaorui Wang and Professor Daisuke Matsushita used existing Osaka City property datasets and incorporated additional information on the physical factors (sky, vegetation, and buildings) of the streetscape images, and the impressions (safety, beauty, depression, liveliness, wealth, and boredom) of the streetscape using machine learning.

The method predicted rent prices with an accuracy of 73.92%. Among the variables, the neighborhood perceptions ranked highly as an indicator, just behind the building age, floor area, and distance to the central business district.

The findings were published in Habitat International.

Funding

This study was supported by JSPS KAKENHI [Grant Number 24K01053], and JST SPRING [Grant Number JPMJSP2139].

Paper information

Journal: Habitat International
Title: Explaining housing rents: A Neural Network approach to landscape image perceptions
DOI: 10.1016/j.habitatint.2024.103250
Authors:  Xiaorui Wang, Jihui Yuan, Yangcheng Gu, Daisuke Matsushita
Published: 30 November 2024
URL: https://doi.org/10.1016/j.habitatint.2024.103250

Contact

Daisuke Matsushita
Graduate School of Human Life and Ecology
Email: matsushita[at]omu.ac.jp

*Please change [at] to @.