Monotonicity of a Bayesian confirmation measure in rule support and confidence
Izabela Brzeziń ska, Roman Słowiń ski · 2005
In knowledge discovery and data mining many measures of interestingness have been proposed in order to reveal different characteristics of the discovered knowledge patterns. Among these measures, an important role is played by Bayesian confirmation measures, which express in what degree a piece of evidence (premise) confirms a hypothesis (conclusion). In this paper, we are considering knowledge patterns in form of “ if… then… ” decision rules with a fixed conclusion. We are investigating the question of monotonic relationship between a particular Bayesian confirmation measure on one side, and rule support and confidence, on the other side. We prove that rules which are optimal according to the confirmation measure, are included in the set of non -dominated rules with respect to both rule support and confidence.