A General Framework for Comparing Heterogeneous Binary Relations

Julien Ah-Pine (1)

1 ERIC - Entrepôts, Représentation et Ingénierie des Connaissances

We propose a general framework for measuring the proximity between heterogeneous binary relations. We are particularly interested in equivalence relations (or partitions, or categorical variables) on the one hand, and in order relations (or rankings or ranked data) on the other hand. We study Kendall's general correlation coefficient that encompasses many statistical association measures for both types of binary relations. Then, we propose a generalization of this approach. Our proposal has several interesting properties. It addresses the case of heterogeneous binary relations, it has a well-motivated geometrical interpretation and it also leads to Euclidean metrics.

Domaines: Statistiques [stat]

Julien Ah-Pine. A General Framework for Comparing Heterogeneous Binary Relations. First International Conference on Geometric Science of Information (GSI 2013), Aug 2013, Paris, France. pp.188-195, ⟨10.1007/978-3-642-40020-9_19⟩. ⟨hal-01504620⟩