Selecting important concepts using weights

Abstract

We present an approach that enables one to select a reasonable small number of possibly important formal concepts from the set of all formal concepts of a given input data. The problem to select a small number of concepts appears in applications of formal concept analysis when the number of all formal concepts of the input data is large. Namely, a user often asks for a list of "important concepts" in such case. In the present approach, attributes of the input data are assigned weights from which values of formal concepts are determined. Formal concepts with larger values are considered more important. The attribute weights are supposed to be set by the users. The approach is a continuation of our previous approaches that utilize background knowledge, i.e. additional knowledge of a user, to select parts of concept lattices. In addition to the approach, we present illustrative examples. © 2011 Springer-Verlag.