In this paper we have proposed PLUMS(Predicting Locations Using Map Similarity), a framework for spatial data mining. We have shown how spatial autocorrelation, the characteristic property of spatial data can be incorporated in the PLUMS framework. When compared with state-of-the-art spatial statistics method in predicting bird-nest locations, PLUMS achieved comparable spatial accuracy while incurring only a fraction of the cost. Futhermore, PLUMS provides a template for specializing other data mining techniques for spatial data.
Our future plan is to bring in other data mining techniques, including
clustering and association rules, within the PLUMS framework. We also
plan to investigate other search algorithms , new map-similarity
measures and non-uniform parameter spaces and determine their dominance
zones.