Edinburgh Napier University

  Formal Concept Analysis often produce huge number of formal concepts even for small input data. Such a large amount of formal concepts, which is intractable to analyze for humans, calls for a kind of
a ranking of formal concepts according to their importance in the given application domain. In this paper, we propose a novel approach to rank formal concepts that utilizes matrix factorization, namely, a mapping of objects and attributes to a common latent space. The lower the distance between objects and/or attributes in the extent and/or intent of a formal concept in the latent space of factors, the more important the formal concept is considered to be. We provide an illustrative example of our approach and examine the impact of various matrix factorization techniques using real-world benchmark data.