Decision Making with Probabilistic Decision Tables

Abstract

The paper is concerned with the decision making with predictive models acquired from data called probabilistic decision tables. The methodology of probabilistic decision tables presented in this article is derived from the theory of rough sets. In this methodology, the probabilistic extension of the original rough set theory, called variable precision model of rough sets, is used. The theory of rough sets is applied to identify dependencies of interest occurring in data. The identified dependencies are represented in the form of a decision table which subsequently is analyzed and optimized using rough sets-based methods. The original model of rough sets is restricted to the analysis of functional, or partial functional dependencies. The variable precision model of rough sets can also be used to identify probabilistic dependencies, allowing for construction of probabilistic predictive models. The main focus of the paper is on decision making aspect of the presented approach, in particular on setting the parameters of the model and on decision strategies to maximize the expected gain from the decisions.