Improving periodic variable star supervised classification using a multistage approach

In astronomy, data analysis paradigm is changing very fast due to the data tsunami that is about to come. Missions like Gaia (launched at December 2013, 50GB/day) or Plato (accepted, 109GB/day) are a good example of the necessity of using and developing automatic data mining algorithms and tools for being able to extract all the possible knowledge from them. In this talk I will present some of the algorithms we have been developing for their integration in the Gaia classification pipeline. In particular, how we can build automatically a multistage classification system to improve the overall classification performance. This advantage is not limited to astronomy, but to any problem with a high number of attributes and classes.