Distributed data mining for astrophysical datasets

Over the past decade, data mining has gained an important role in astronomical data analysis. Traditionally, such analysis is performed on data at a single location. However, one of the main motivational forces behind a virtual observatory is the distributed nature of both data and computational resources. Existing data-mining methods for distributed data are either communication-intensive or result in a loss of accuracy. In this paper, we introduce a general approach to supervised data mining that allows data to remain distributed, but still produces satisfactory results. We demonstrate by applying the approach to a number of astronomical datasets.