Mining multivariate associations within GIS environments

Abstract

As geospatial data grows explosively, needs for the incorporation of data mining techniques into Geographic Information Systems (GISs) are in great demand. Association rules mining is a core technique in data mining and is a solid candidate for the cause-effect analysis of large geospatial databases. It efficiently detects frequent asymmetric causal patterns in large databases. In this paper, we investigate a series of geospatial preprocessing steps involving data conversion and classification so that traditional boolean and quantitative association rules mining can be applied. We present a robust geospatial multivariate association rules mining framework for efficient knowledge discovery within data-rich GISs environments. The proposed approach can be integrated into traditional GISs using dynamic link library and scripting languages such as AVENUE for ArcView and MapBasic for MapInfo. Our framework is designed and implemented in AVENUE for ArcView GIS. Experiments with real datasets demonstrate the robustness and efficiency of our approach.