Spatial data discretization methods for geocomputation

Abstract Geocomputation provides solutions to complex geographic problems. Continuous and discrete spatial data are involved in the geocomputational process; however, geocomputational methods for discrete spatial data cannot be directly applied to continuous or mixed spatial data. Therefore, discretization methods for continuous or mixed spatial data are involved in the process. Since spatial data has spatial features, such as association, heterogeneity and spatial structure, these features cannot be handled by traditional discretization methods. Therefore, this work develops feature-based spatial data discretization methods that achieve optimal discretization results for spatial data using spatial information implicit in those features. Two discretization methods considering the features of spatial data are presented. One is an unsupervised method considering autocorrelation of spatial data and the other is a supervised method considering spatial heterogeneity. Discretization processes of the two methods are exemplified using neural tube defects (NTD) for Heshun County in Shanxi Province, China. Effectiveness is also assessed.