An extended formalism to constraint logic programming for decision analysis

摘要

While constraint logic programming (CLP) is becoming a favorite tool for decision support systems (DSS), its utility to DSS is limited due to the lack of decision theoretic analysis capability. The combination of CLP with decision theoretic analysis is therefore necessary for more fruitful CLP applications to DSS. In this paper, we propose a framework which embeds decision tree analysis into CLP. The framework provides an integrated representation of decision problems with logic, constraints, probability, and utility. The benefits are threefold. The first is to offer an integrated representation and reasoning mechanism for general knowledge-based decision analysis problems. The second is to provide support to automatic or computer-aided construction of decision trees from the declarative representation of decision problems so that a decision maker can get an intuitive decision scenario from decision trees without manual labor. Last of all, the mechanism of constraint propagation provided by CLP significantly reduces the complexity of decision trees by removing infeasible solutions.