Mining confident rules without support requirement

Abstract

An open problem is to find all rules that satisfy a minimum confidence but not necessarily a minimum support. Without the support requirement, the classic support-based pruning strategy is inapplicable. The problem demands a confidence-based pruning strategy. In particular, the following monotonicity of confidence, called the universal-existential upward closure, holds: if a rule of size k is confident (for the given minimum confidence), for every other attribute not in the rule, some specialization of size k+1 using the attribute must be confident. Like the support-based pruning, the bottleneck is at the memory that often is too small to store the candidates required for search. We implement this strategy on disk and study its performance.