Rental prices are typically influenced by tangible factors such as the age of the building, available amenities, and location. However, not all properties with similar physical attributes command the same rent. The subjective perceptions of a neighbourhood also play a crucial role.

A team from Osaka Metropolitan University has honed a technique that accounts for these perceptual variables, achieving nearly 75% accuracy in predicting housing prices within Osaka City.

Xiaorui Wang, a student at the Graduate School of Human Life and Ecology, and Professor Daisuke Matsushita spearheaded this initiative. They leveraged existing property datasets from Osaka City, enhancing them with additional data concerning physical aspects of street-level images (such as the sky, vegetation, and architecture). Furthermore, they assessed the streetscape based on various impressions — safety, beauty, depression, liveliness, wealth, and boredom — using machine learning techniques.

Their methodology successfully predicted rental prices with an accuracy of 73.92%. Notably, the perceptions of the neighbourhood emerged as a significant factor, ranking just below the building’s age, its floor area, and its proximity to the central business district in terms of influence on rental pricing.

More information: Xiaorui Wang et al, Explaining housing rents: A neural network approach to landscape image perceptions, Habitat International. DOI: 10.1016/j.habitatint.2024.103250

Journal information: Habitat International Provided by Osaka Metropolitan University