An Extended Semi-Supervised Regression Approach with Co-Training and Geographical Weighted Regression: A Case Study of Housing Prices in Beijing

ISPRS International Journal of Geo-Information

Yi Yang¹  Jiping Liu²  Shenghua Xu²  Yangyang Zhao² 

[1] School of Resource and Environmental Science, Wuhan University, No. 129 Luoyu Road, Wuhan 430079, ChinaResearch Center of Government GIS, Chinese Academy of Surveying and Mapping, No. 28 Lianhuachi West Road, Haidian District, Beijing100830, China;

关键词: semi-supervised regression;    geographical weighted regression;    spatial nonstationarity;    housing prices;   

DOI  :  10.3390/ijgi5010004

来源: mdpi

【 摘 要 】

This paper proposes an extended semi-supervised regression approach to enhance the prediction accuracy of housing prices within the geographical information science field. The method, referred to as co-training geographical weighted regression (COGWR), aims to fully utilize the positive aspects of both the geographical weighted regression (GWR) method and the semi-supervised learning paradigm. Housing prices in Beijing are assessed to validate the feasibility of the proposed model. The COGWR model demonstrated a better goodness-of-fit than the GWR when housing price data were limited because a COGWR is able to effectively absorb no-price data with explanatory variables into its learning by considering spatial variations and nonstationarity that may introduce significant biases into housing prices. This result demonstrates that a semisupervised geographic weighted regression may be effectively used to predict housing prices.

【 授权许可】

CC BY   
© 2016 by the authors; licensee MDPI, Basel, Switzerland.