An Empirical Study of Observation-weighting Method for Mining Actionable Behavioral Rules
Peng Su, Yufeng Wang · Journal of Physics Conference Series · 2020
Abstract Knowledge is considered actionable if users can take direct actions based on such knowledge to their advantage. Among a variety types of actionable knowledge, the actionable behavioral rule plays an important and unique role because it can directly and explicitly suggest actions to users to positively influence their behaviors of concerned entities. The problem of mining such rules is a search problem within a framework of support and expected utility. The previous mining approach assumes that each instance which supports a rule has the uniform contribution to the support. However, this assumption is usually violated in practice, and thus will hinder the performance of algorithms for mining such rules. To handle this problem, in this paper, we propose several observation-weighting models for support based on different functions. We further empirically investigate these models. Based on the results of our experimental study, we gain a thorough insight into the selection of various observation-weighting models for this domain.