多源信息的验前分布融合方法研究

验前分布的获取和表示是B ayes理论应用中的一个关键问题。针对目前多源验前信息融合中存在的过分依赖专家信息的问题,提出了一种基于第2类极大似然估计原理(M L-II)确定验前分布权重的方法,该方法将现场试验数据看作是由其边缘分布产生的样本,然后根据不同验前分布下现场样本似然性的大小来确定其在验前分布融合中的作用,并以此进一步确定不同验前分布的权重因子。最后通过算例证明了该方法比基于专家信息的融合方法更为合理和有效。

Research on Fusion Method for Prior Distribution of Multiple Sources Information

How to acquire proper prior distribution is a key problem in Bayesian analysis.In this paper,the situation is that the fusion of information of multiple sources relies on expert’s information excessively for the moment.Based on the second maximum likelihood estimation theory,a new method which is used to confirm the weight of the prior distribution is proposed.According to the method,the field test data is seen as the sample which is generated by the marginal distribution.The weight factors of different prior distributions are confirmed by the magnitude of likelihood of the field sample which appears under different prior distributions.In the end,an example shows that the method proposed is more reasonable and more effective than the fusion method based on the expert’s information.