Spatial Congeries Pattern Mining

Speaker: Iris ZHANG

Abstract

Spatial databases are databases which store tremendous amounts of spatial and non-spatial data. They don't store explicitly implicit regulations, rules and patterns which are useful for environmental research, government layout and geo-marketing. Knowledge discovery in large spatial databases, or spatial data mining, is to extract such kinds of relations or patterns which don't appear in the databases. Recently, a new research area in spatial data mining has developed. It groups data points according to their spatial attributes and then finds patterns of their non-spatial attributes. For example, “the areas which often have level-5 forest fire, the area which are very drought and the areas which have little precipitation, often intersection” is such a pattern. Such kind of patterns can be used in applications in ecology and environment management, public safety, and e-business. In this seminar, representative techniques, neighboring class sets mining and co-location pattern mining, will be introduced. A new technique “spatial congeries pattern mining”, which can make up some shortcomings of these two techniques, is defined. The talk will give a formal definition of spatial congeries patterns and introduce some mining algorithms for them. Some preliminary experimental results will also be presented.

Referred Papers