Parametrised measures for the evaluation of association rule interestingness

Stéphane Lallich (1) , Benoît Vaillant (2, 3) , Philippe Lenca (2, 3)

1 ERIC - Equipe de Recherche en Ingénierie des Connaissances
2 LUSSI - Département Logique des Usages, Sciences sociales et Sciences de l'Information
3 TAMCIC - Traitement Algorithmique et Matériel de la Communication, de l'Information et de la Connaissance

In this paper, we present a original and synthetical overview of most commonly used association rule interestingness measures. These measure usually relate the confidence of a rule to an independency reference situation. Others relate it to indetermination, or impose a minimum confidence threshold. We propose a systematic generalisation of these measures, taking into account the reference point choosen by an expert in order to apprehend the confidence of a rule. This generalisation introduces new connections between measures, leads to the enhancement of some of them, and we propose new parametrised possibilities.

Mots clés: Measures

Domaines: Ordinateur et société [cs.CY]

Stéphane Lallich, Benoît Vaillant, Philippe Lenca. Parametrised measures for the evaluation of association rule interestingness. ASMDA 2005 : XIth International Symposium on Applied Stochastic Models and Data Analysis, May 17-20 Brest, France, May 2005, Brest, France. pp.220 - 229. ⟨hal-02124614⟩