基于软概率的最小风险贝叶斯决策

Minimum Risk Bayes Decision-making Based on Soft Probability

为有效处理限制样本、非随机稳定信息下的不确定型决策问题,文章提出了一种基于软概率的最小风险贝叶斯决策方法:确定先验信息中各状态的软概率区间值,利用区间概率下的贝叶斯风险模型输出各后验状态的区间值决策结果,利用基于可信度的区间数排序方法对结果按照风险大小进行排序,从而实现决策风险最小化的后验状态判定,同时,随着新样本信息的加入,决策结果会相应动态调整,更趋确定。同时,用案例验证了该方法的合理性和可行性。

In order to effectively deal with uncertain decision-making under limited samples and non-stochastic stable information, this paper proposes a minimum risk Bayes decision-making method based on soft probability: The soft probability interval values of each state in the prior information are determined, and the interval decision results of each posterior state are output by the Bayesian risk model under the interval probability. The interval number ranking method based on confidence is used to rank the results according to the risk, so as to realize the posterior state judgment of decision-making risk minimization. At the same time, with the addition of new sample information, the decision results will be dynamically adjusted and more likely to be determined. The paper uses a case to verify that the method is reasonable and feasible.