EUVIMEDEUVIMED entdecken
Crossref

Residential Price Forecasting in Shaoxing using Gaussian Process Regression with Bayesian Optimisation

Bingzi Jin, Xiaojie Xu

International Journal of Real Estate Studies · 30.06.2026 · 20 (1) · 5-21

DOI: 10.11113/intrest.v20n1.436 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Real Estate Studies, online erschienen am 30.06.2026. Verfasst von Bingzi Jin und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The Chinese residential property market has undergone pronounced cyclical shifts over the last decade, moving from sustained expansion to a sharp downturn after late 2021, which poses considerable challenges for investors, planners, and regulators seeking to anticipate urban housing price trends. While machine learning methods are increasingly applied to real estate forecasting, Gaussian process regression (GPR) remains underutilised in China’s spatially heterogeneous markets, and no prior study has deployed it to model price dynamics in Shaoxing—a representative, rapidly transitioning urban centre. To address this gap, the present study constructs a GPR forecasting framework with multiple kernel structures and adaptive basis functions, optimized through Bayesian inference and cross-validation, using monthly residential price data from January 2013 to July 2024. The model is trained on the period up to April 2022 and evaluated out-of-sample over May 2022–July 2024. For comparative benchmarking, identical test conditions are applied to a long short-term memory (LSTM) network, support vector regression, a regression tree, and a simple autoregressive model. The GPR model achieves a root mean square error of 30.98, markedly lower than LSTM (48.58), support vector regression (60.98), regression tree (73.34), and the autoregressive benchmark (87.28), corresponding to a relative root mean square error of 0.1774%. These results confirm the framework’s ability to capture the nonlinear temporal dependencies that characterise Shaoxing’s housing market. Beyond its methodological contribution, the study offers practical value for valuation practice, land administration, and housing policy by providing a transparent, scalable, and data-driven tool for price monitoring. The proposed system can be deployed independently or integrated with conventional econometric models, and its parsimonious input design facilitates replication in other urban contexts, supporting evidence-based decision-making across diverse real estate markets.

Zitieren

APA

Jin, B., & Xu, X. (2026). Residential Price Forecasting in Shaoxing using Gaussian Process Regression with Bayesian Optimisation. International Journal of Real Estate Studies, 20(1), 5-21. https://doi.org/10.11113/intrest.v20n1.436

Vancouver

Jin B, Xu X. Residential Price Forecasting in Shaoxing using Gaussian Process Regression with Bayesian Optimisation. International Journal of Real Estate Studies. 2026;20(1):5-21. doi:10.11113/intrest.v20n1.436

BibTeX
@article{Jin2026,
  author = {Jin, Bingzi and Xu, Xiaojie},
  title = {Residential Price Forecasting in Shaoxing using Gaussian Process Regression with Bayesian Optimisation},
  journal = {International Journal of Real Estate Studies},
  year = {2026},
  volume = {20},
  number = {1},
  pages = {5--21},
  doi = {10.11113/intrest.v20n1.436}
}
Kostenlos starten →