Abstract
Property valuation can reflect the external influences of environmental and locational attributes, allowing for adjustments to land development and urban planning policies accordingly. In a dense city setting with substantial amounts of high-rise buildings, the current property valuation needs updates regarding concepts, data, and technology to add the vertical dimension (3D), namely 3D property valuation (3DPV). Previous studies have extensively analysed how different locational and environmental attributes influence property values within a hedonic price model (HPM) on a horizontal (2D) basis. However, little consideration is given to 3D. As the cities grow vertically to shelter the growing urban population, the spatial changes in the 3D built environment and their influences on property values should be paid more attention to. To this end, the main research objective of this thesis is to investigate the vertical dimension (3D) in property valuation using 3D modelling and machine learning (ML) within a dense city setting, such as, Xi’an, China. Xi’an is a provincial city with a population of over 10 million and an important political and economic role in China, which makes it an appropriate study area. The first objective is to identify the requirements of experts and end-users for adding 3D data in property valuation in a complex urban area. This objective was achieved by a qualitative study conducted in Xi’an, China, which collected the opinions of different stakeholders regarding 3D through semi-structured expert interviews, focus group discussions, and questionnaires. It was identified that real estate developers have a profit-driven attitude; the local government has not yet considered the 3D aspect in the housing policy; buyers already find the 3D variables very beneficial in terms of living quality. Buyers assign 3D variables an overall higher score than 2D variables. Based on the discussions, it was concluded that it would be of great value to explore the nexus of how 3D modelling, a technical method, can make 3D more visible in property valuation. Secondly, after setting up the conceptual base, this research aims to identify the spatial analytical capabilities of 3D modelling and the importance of 3D variables in property valuation. HPMs were constructed with 2D and 3D variables, respectively. The 2D HPM results show that a highly regulated property market cannot reflect the influences of 2D variables; comparatively, the 3D HPM explains the property value variation at the neighbourhood scale better than 2D models. This study provides insights into the importance of 3D variables in HPM and the spatial analytical capabilities of 3D modelling at the neighbourhood scale. 199 Summary However, the prime weakness is transferring and extending to a city-scale mass study. After showing the societal needs as well as the challenges to including 3D in property valuation, a state-of-the-art review of urban 3D modelling methods related to property valuation was conducted due to the lack of ready-to-use 3D modelling methods for 3DPV. Therefore, a review of urban 3D modelling methods and data was conducted, in which these methods and data are assessed regarding their adaptability for 3DPV according to a set of self-developed analytical criteria. The overarching methodology is desk-based, including systematic literature review and snowball sampling in literature search and semi-structured interviews to develop a set of analytical criteria. This review also analysed the state-of-theart of 3DPV and put forward 3DPV prospects in the short and long term, which can be a base for future 3DPV-related works. Finally, with all the knowledge and practices gained previously, the final empirical study proposed a novel method using Street View Images (SVI) acquired from two angles, eye-level (pitch 0°) and sky-view (pitch 90°, upwards), to bring the 3D perspective in, and ML for creating a city-scale 3DPV model. It extended the classic HPM by adding 3D variables extracted from SVI. The main findings firstly proved the statistical significance of 3D variables in HPM; secondly, the impacts of different 3D variables on property values were identified (e.g., in the sky-view angle, the proportion of sky has a positive correlation while the presence of buildings and trees are negatively correlated with property values); third, Random Forest (RF), the specific ML algorithm used in this study, outperforms other statistical models with the highest R2 (0.768) and the least rootmean-square deviation (1,669.60 CNY/m2). The relative feature importance and partial dependence plots diagrams provided by RF help us understand how and when 3D variables influence property values. This study provides an innovative way to use SVI in different pitches to represent the 3D built environment in valuation models. Returning to the main research objective - to investigate the vertical dimension (3D) in property valuation using 3D modelling and machine learning (ML) in dense cities – it becomes evident that the lack of a 3D perspective in a dense city setting limits the exploration of 3D in valuation practices. The results of this thesis contribute to the existing literature by broadening the understanding of the relationship between the 3D built environment and property values. Via qualitative and quantitative studies and a comprehensive review, the findings show that (1) 3D should be added to property valuation in dense cities; (2) the current urban 3D modelling methods and data support need more adaptation for 200 Summary property valuation purposes; and (3) SVI and ML benefit 3DPV regarding model explainability and 3D geoinformation provision. Although this thesis focuses on manifesting the role of 3D in property values, the research results can provide references for other applications and societal questions, such as social segregation and gentrification, building auditing, property taxation and landscape design.
Original language | English |
---|---|
Qualification | Doctor of Philosophy |
Awarding Institution |
|
Supervisors/Advisors |
|
Award date | 17 Oct 2024 |
Place of Publication | Enschede |
Publisher | |
Print ISBNs | 978-90-365-6250-8 |
Electronic ISBNs | 978-90-365-6251-5 |
DOIs | |
Publication status | Published - 17 Oct 2024 |