Data Mining in Astronomy: Classification of Eclipsing Binaries

摘要

Data mining is a powerful tool to obtain new knowledge and make scientific discoveries. One of key problems in astronomy is the classification of astronomical objects based on their observational parameters. Main goal of the current presentation is a description of method and results of data mining application to automatic classification of eclipsing binaries. The method is based on the data from a thousand classified systems and allows for the classification of a given system based on a set of observational parameters, even if the set is incomplete. The procedure is applied to large catalogues of eclipsing variables, including those obtained as by-products of microlensing surveys (OGLE, MACHO, ASAS-3). Also, after careful analysis of Data Mining methods and approaches of their incorporation into the Virtual Observatory infrastructure (AstroGrid) the CEA-application has been developed under the name Ensembled Weka that provides for various variants of composition of ensembles of algorithms. Ensembled Weka solves a problem defined by a problem description file. The design gives also an ability of hierarchical data analysis. Ensembled Weka has been checked by solving of eclipsing binaries classification problem.