Generalization of association rules through disjunction

Abstract

Several efforts were devoted to mining association rules having conjunction of items in premise and conclusion parts. Such rules convey information about the co-occurrence relations between items. However, other links amongst itemslike complementary occurrence of items, absence of items, etc.-may occur and offer interesting knowledge to end-users. In this respect, looking for such relationship is a real challenge since not based on the conjunctive patterns. Indeed, catching such links requires obtaining semantically richer association rules, the generalized ones. These latter rules generalize classic ones to also offer disjunction and negation connectors between items, in addition to the conjunctive one. For this purpose, we propose in this paper a complete process for mining generalized association rules starting from an extraction context. Our experimental study stressing on the mining performances as well as the quantitative aspect proves the soundness of our proposal.

Construction efficace du treillis des motifs fermés fréquents et extraction simultanée des bases génériques de règles

“…Les perspectives de travaux futurs concernent les pistes suivantes : -Du point de vue de la qualité des connaissances extraites par PRINCE, l'adaptation de ce dernier pour l'exploration de l'espace disjonctif de recherche, où les items sont liés par l'opérateur de disjonction au lieu de l'opérateur de conjonction [Hamrouni & al., 2009], est une piste intéressante afin d'extraire différentes formes de règles généralisées [Hamrouni & al., 2010]. Ces dernières sont connues pour être plus riches en information étant donné qu'elles véhiculent différents liens entre les items et ne se limitent donc pas à leur conjonction.…”

Expert deduction rules in data mining with association rules: new results and challenges

“…As stated above, it follows from (14), that for each α > 0, there are n a ∈ N and n b ∈ N, such that (20) holds. Thus, if we set n = max{n a , n b }, then (21) holds. This finishes the proof of (15).…”

Interpretable Patient Mortality Prediction with Multi-value Rule Sets

“…For example, a 1 ∨ a 2 ∨ • • • → Y , where a 1 is an item (e.g., gender = female). Representative works include [4][5][6]. is primitive form of rules was extended to consider disjunctions in the condition/literal level [12], yielding multi-value rules of the form (a…”