Using GIS in the Development of Forecasting Models

Publication Date


Committee Chair

Karen K. Kemp, Ph.D.

Committee Members

Avijit Sarkar, Ph.D.


Using Geographic Information Systems (GIS) adds a new perspective to residential house price prediction. The use of address level data for house sales locations combined with demographic data from the census and data describing the proximity (Euclidean Distance) of those houses to features viewed as cost criteria greatly enhances the accuracy of forecasts using parametric and nonparametric methods. This project supports work completed by Dr. Mak Kaboudan (Professor of Business Statistics at the School of Business, University of Redlands) to develop house price forecast models for the City of Redlands.

These models were developed to improve on previous models that estimate single family house prices through the use of GIS. It differs from other models by using a Location Price Index (LPI) as a generalization of values of residential houses sold in a particular neighborhood. LPI is a ratio of neighborhood prices relative to city home prices developed to capture the impact of variations in average neighborhood house attributes-- structure, location, and socioeconomic differences--on home valuations. After the model results are reported, a spatial assessment was performed to investigate alternative methodologies that could be used in future studies.

The focus of this study is not on the models’ accuracy, but on the use of GIS and spatial analysis as tools in the development and evaluation of forecasting models. This study is designed to identify how GIS and spatial analysis can be used to help models acquire more accurate results. It discusses how spatial analysis prior to model development can help to identify problems that may arise in model development. Spatial analysis can also be used to identify variables that have high correlation with the dependent variable to be forecasted.

Full text is available at the University of Redlands


Article Location