Dynamic focus programming: A new approach to sequential decision problems under uncertainty
Dynamic focus programming: A new approach to sequential decision problems under uncertainty
复制标题
DOI：
10.1016/j.ejor.2022.02.044
复制
发表时间：
2022-03
期刊：
影响因子：
--
通讯作者：
P. Guo
中科院分区：
文献类型：
--
作者：
P. Guo
利用焦点选择理论，提出了一种解决不确定性下顺序决策问题的新方法，即动态焦点规划。在动态焦点编程中，有两种不同的评估系统：积极的和消极的。检查由从初始阶段到最终阶段的决策序列以及相关状态组成的每个可能路径。在正向评价体系中，对于初始阶段的每个决策，如果从该决策出发的一条路径能够以较高的概率带来较低的总成本，则选择该路径作为该决策的正向焦点路径；基于所有初始决策的积极焦点路径，决策者选择最优选的决策规则。在负面评价体系中，对于初始阶段的每个决策，如果从该决策出发的一条路径能够以较高的概率带来较高的总成本，则选择该路径作为该决策的负面焦点路径；基于所有初始决策的负面焦点路径，决策者选择最可接受的决策规则。对于特定的顺序决策问题，仅激活一个系统；至于哪一个起作用，很大程度上取决于决策者的个性和框架。我们将动态焦点规划应用于实际的投标决策问题：我们获得最优决策规则并获得决策者的行为洞察。
A new approach to sequential decision problems under uncertainty named dynamic focus programming is proposed with the focus theory of choice. In dynamic focus programming, there are two distinct evaluation systems: Positive and negative ones. Each possible path consisting of a decision sequence from the initial stage to the final stage and the associated states is examined. In the positive evaluation system, for each decision in the initial stage, if a path starting from it can bring about a relatively low total cost with a relatively high probability, then this path is selected as the positive focus path of this decision; based on the positive focus paths of all initial decisions, a decision maker chooses a most-preferred decision rule. In the negative evaluation system, for each decision in the initial stage, if a path starting from it can bring about a relatively high total cost with a relatively high probability, then this path is selected as the negative focus path of this decision; based on the negative focus paths of all initial decisions, a decision maker chooses a most acceptable decision rule. For a specific sequential decision problem, only one system is activated; as for which one works, it is strongly dependent on decision maker's personality and the framing. We apply dynamic focus programming to a real bidding decision-making problem: We obtain the optimal decision rule and gain the behavioral insights of the decision maker.