A rule-based risk decision-making approach and its application in China's customs inspection decision

This paper addresses a kind of risk decision-making problem existing widely in public administration and business management, which is characterized by (1) occurrence probabilities of states of nature can be estimated by analysing historical observations, but historical observations of different objects are unhomogeneous, (2) the relation between observations and occurrence probabilities of states of nature are affected by some qualitative and quantitative indicators, (3) it is a real-time decision-making problem, that is, there are many decisions for different objects to be made in a limited time, (4) considering decision's execution, impact of resource constraints is an important issue in decision-making process. In this paper, we develop a rule-based approach to address the problem. In the proposed approach, a two-step clustering method is employed to classify objects into categories, and observations in each category can be approximately viewed as homogeneous. For objects in each category, occurrence probabilities of states of nature are estimated by logistic regression, and the decision rule is obtained through solving an optimization model, which is to minimize the total decision risks while satisfying resource constraints. Effect and efficacy of our approach are illustrated through its application to China's customs inspection decision.