Using geographical information system - multiple regression analysis - generated location value response surface approach to model locational factor in the prediction of residential property values

• Main Author: Chin, Chui Vui
• Format: Thesis
• Language: English
• Published: 2006
• Subjects: GA Mathematical geography. Cartography
• Online Access: http://eprints.utm.my/id/eprint/5144/1/ChinChuiVuiMFKSG2006.pdf; http://eprints.utm.my/id/eprint/5144/

The limitation and highly complex process of discrete measurement of location have encouraged searching for alternative approach to derive locational compensation factor in the prediction of property values. A new approach by use of Geographical Information System-Multiple Regression Analysis-generated location value response surface (GIS-MRA-generated LVRS) approach was proposed in this study. The method used LVRS in a GIS to model locational influence from MRAgenerated residuals, with no locational variables on residential property values in the local context by integrating spatial & aspatial data in term of developing a hybrid predictive model. This study has three main objectives. First, to discuss the pertinent factors influencing residential property values, including the location factors. Second, to develop a predictive model of residential property values, whereby the locational factor is modelled by GIS-MRA-generated LVRS. Third, to examine the usefulness of GIS-MRA-generated LVRS to improve the quality of the regression model. To achieve these objectives, this study was divided into two main parts. The first part comprised the theories of value, residential property value factors, location modelling, MRA modelling, and the spatial interpolation techniques. The second part, comprised the development of hybrid models, whereby the locational factors was modelled by GIS-MRA-generated LVRS approach and examining its ability to improve regression modelling. As many as125 individual terraced units in three adjoining housing schemes (Taman Pelangi, Taman Sentosa and Taman Sri Tebrau) in Johor Bahru were used for model estimation, while 14 transacted units were set aside for predictive purposes. Results have shown models applying LVRS have managed to improve overall model’s statistical quality and predictive performance by achieving higher proportion of “reasonably accurate� prediction as compared to the traditional MRA models. The LVRS has allowed a clearer spatial visual picture of the location influence to be captured at all level in the study area. The location factor influence to property value can be modelled in a more effective way.