Developing a decision tree framework for mining spatial association patterns from GIS database

Abstract

Spatial data mining and knowledge discovery (SDMKD) is a whole process of discovering implicit but useful knowledge from GIS databases. From the first law of geography, spatial association patterns are the realizations of processes that operate across the geographic space. This paper attempts to present a decision tree framework to assist in analyzing spatial association patterns. Based on the problem, the representation of data or data model should be identified firstly. Secondly, geostatistical, lattice and point pattern data can be distinguished through the characteristics of spatial domain. The main task of third level of the decision tree is to apply different spatial data analysis methods to different spatial data types. For lattice data, the work is to apply exploratory spatial data analysis (ESDA) to find spatial association patterns, and then identify the driving forces which cause the observed spatial association patterns by confirmatory spatial data analysis (CSDA). The fourth level is to verify the precision and accuracy of spatial association models. All in all, spatial association pattern analysis is a process of acquiring useful spatial patterns by circulation and repetition.