A case-based reasoning approach for land use change prediction

Abstract

Although has been widely used to study geographical problems, case-based reasoning (CBR) method is far less than perfect and research is in great need of to improve CBR-based geographic data representation modeling, as well as spatial similarity computation and reasoning algorithm. This paper reports an improved CBR-based method for studying the spatially complex land use change. Based on a brief summary of advantages and challenges of current existing quantitative methods, the paper first proposes to introduce the CBR approach for land use change study. A three-component model (“problem”, “geographic environment”, and “outcome”) was proposed to represent the land use change cases among which there are complicated and inherent spatial relationships. This paper then presents an algorithm to retrieve the inherent spatial relationships, which are then introduced into the CBR similarity reasoning algorithm to predict land use change. The method was tested by examining the land use change in Pearl River Mouth area in China and yields a similar prediction accuracy of 80% as that derived by applying the Bayesian networks approach to the exact same data. As a result, the CBR-based method proposed in this study provides an effective and explicit solution to represent and solve the complicated geographic problems.