Forecasting spatial dynamics of the housing market using Support Vector Machine

This paper adopts a novel approach of Support Vector Machine (SVM) to forecast residential housing prices. as one type of machine learning algorithm, the proposed SVM encompasses a larger set of variables that are recognized as price-influencing and meanwhile enables recognizing the geographical pat...

Full description

Saved in:
Bibliographic Details
Published in:International journal of strategic property management 2017-09, Vol.21 (3), p.273-283
Main Authors: Chen, Jieh-Haur, Ong, Chuan Fan, Zheng, Linzi, Hsu, Shu-Chien
Format: Article
Language:English
Subjects:
Data mining
Dwellings
Economy
Financial Markets
Forecasting
Forecasts and trends
Hedonic appraisal method
Housing
Housing price forecasting
Housing prices
Machine learning
Rural and urban sociology
Spatial dynamics
Support vector machines
Supporting vector machine
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper adopts a novel approach of Support Vector Machine (SVM) to forecast residential housing prices. as one type of machine learning algorithm, the proposed SVM encompasses a larger set of variables that are recognized as price-influencing and meanwhile enables recognizing the geographical pattern of housing price dynamics. The analytical framework consists of two steps. The first step is to identify the supporting vectors (SVs) to price variances using the stepwise multi-regression approach; and then it is to forecast the housing price variances by employing the SVs identified by the first step as well as other variables postulated by the hedonic price theory, where the housing prices in Taipei City are empirically examined to verify the designed framework. Results computed by nonparametric estimation confirm that the prediction power of using SVM in housing price forecasting is of high accuracy. Further studies are suggested to extract the geographical weights using kernel density estimates to reflect price responses to local quantiles of hedonic attributes.
ISSN:1648-715X
1648-9179
DOI:10.3846/1648715X.2016.1259190