TY - JOUR
T1 - The Spatial neural network model with disruptive technology for property appraisal in real estate industry
AU - Lin, Regina Fang Ying
AU - Ou, Chiye
AU - Tseng, Kuo Kun
AU - Bowen, Deng
AU - Yung, K. L.
AU - Ip, W. H.
N1 - Funding Information:
We thank Shenzhen Government for generously providing the funding for this research [JCYJ20190806144609107, HA11409015, and GXWD20201230155427003 - 20,200,829,144,221,001]. We also thank the anonymous referees for their comments and guidance. All errors are ours.
Publisher Copyright:
© 2021
PY - 2021/12
Y1 - 2021/12
N2 - Property valuation is a complex issue that has always been the focal point for the real estate industry. The traditional valuation models used for appraisals cannot meet real-world demand anymore due to the improper processing of correlated information of nearby facilities. In this study, we propose a Spatial Neural Network (SNN) model, called Property Appraisal 4.0, that uses disruptive technology to forecast property values and discover hidden neighbourhood features of real estate information in the satellite embedding vectors. The latest deep learning technologies are also employed, such as knowledge distillation, incremental learning, and Deep-Automated Optical Inspection. Class Activation Mapping is also adapted to reinforce the proposed spatial neural network in the model. Experimental results show that our approach's performance is better than that of previous mainstream models, such as the Hedonic Pricing Model and Support Vector Machines.
AB - Property valuation is a complex issue that has always been the focal point for the real estate industry. The traditional valuation models used for appraisals cannot meet real-world demand anymore due to the improper processing of correlated information of nearby facilities. In this study, we propose a Spatial Neural Network (SNN) model, called Property Appraisal 4.0, that uses disruptive technology to forecast property values and discover hidden neighbourhood features of real estate information in the satellite embedding vectors. The latest deep learning technologies are also employed, such as knowledge distillation, incremental learning, and Deep-Automated Optical Inspection. Class Activation Mapping is also adapted to reinforce the proposed spatial neural network in the model. Experimental results show that our approach's performance is better than that of previous mainstream models, such as the Hedonic Pricing Model and Support Vector Machines.
KW - Class activation mapping
KW - Deep-Automated Optical Inspection (AOI)
KW - Disruptive technology
KW - Real estate valuation
KW - Spatial information
KW - Spatial neural network
UR - http://www.scopus.com/inward/record.url?scp=85113306860&partnerID=8YFLogxK
U2 - 10.1016/j.techfore.2021.121067
DO - 10.1016/j.techfore.2021.121067
M3 - Journal article
AN - SCOPUS:85113306860
SN - 0040-1625
VL - 173
JO - Technological Forecasting and Social Change
JF - Technological Forecasting and Social Change
M1 - 121067
ER -