Extending Redescription Mining to Multiple Views

Abstract

Redescription mining is a data mining task that discovers re-descriptions of different subsets of entities from available data. Locating such re-descriptions is important in many scientific disciplines because it allows detecting different types of associations including synergy of different attributes of interest. There exist a number of redescription mining algorithms, however they are all restricted to use of one or maximally two disjoint sets of attributes (views) to re-describe different subsets of entities. The main reasons for this limitation are computational complexity and potentially large increase in number of produced patterns, in multi-view setting, during redescription mining. In this work we present an algorithm that allows mining redescriptions from multiple views using the CLUS-RM algorithm. Presented algorithm efficiently solves aforementioned problems. Its computational complexity, with respect to attribute operations, increases linearly with the increase of number of views and we present techniques to handle large number of produced redescriptions during redescription mining step.