Clustering and Overlay Analysis for Discovering Spatial Association Rules

An explosion of geographic data sets and related processing has occurred in recent years. Spatial data mining techniques are of interest to handle these large volumes of data and to identify relationships between features in a data set. Data mining holds the promise of being able to discover and validate association rules, to quickly determine whether a feature of interest is present, and to predict the existence of missing features. A spatial data mining technique is proposed which can work with data from multiple sources, can handle both numeric and categorical data, and can account for spatial neighborhood effects. This technique uses clustering, overlay, and statistical analyses to determine whether a group of features or attribute values are likely to appear together in a data set, and if so, whether an association rule can be validated between these features. The method will create a representational format for the data, identify both vertical and horizontal relationship types and determine their probability, and test and validate the specific relationships found. The Phase I project will demonstrate the feasibility of this approach, with the purpose of predicting missing features or rapidly identifying significant features in a large data set.