摘要
为了消除和减弱当证据层不满足条件独立性假设时对预测结果产生的影响,提出了逐步证据权模型和加权证据权模型.加权证据权模型通过对logit模型进行修改,对各个证据层给予一定的权重,以调整由于证据层与其他证据层的条件相关性对模型的影响;逐步证据权模型是将证据层按照一定的顺序逐步加入到模型中,在加入到模型的过程中依次用已经获得的后验概率作为模糊训练层的方法.以个旧锡铜多金属矿产资源预测为例,应用4种证据权模型的后验概率进行异常圈定,结果表明两种新的模型对减弱证据层不满足条件独立性假设所产生的影响是有效的.
This paper proposes two kinds of new models of weight of evidence： weighted weights of evidence and stepwise weights of evidence to reduce the influence of correlation among evidence layers when hypothesis of conditional independence is not held. In the weighted weights of evidence model, an adjustment of weight of evidence is made to reduce the influence of correlation among evidence layers. In the stepwise model, each evidence layer is added into the logistic model as if a single layer whose weight can be calculated by using the previously calculated posterior probability as a new prior probability. These two models are compared with other models through a case study of calculating posterior probability maps for Sn-Cu mineral deposits in Gejiu, Yunnan, China. The result shows that both models are effective to reduce the influence of correlations among evidence layers on delineation of Sn-Cu anomalies.
出处
《地球科学（中国地质大学学报）》
EI
CAS
CSCD
北大核心
2009年第2期281-286,共6页
Earth Science-Journal of China University of Geosciences
基金
国家自然科学基金重点项目(No.40638041)
地质调查项目(No.121201063390110)
地质过程与矿产资源国家重点实验室开放课题(No.GPMR200803)
国家863项目(Nos.2006AA06Z115,2006AA06Z113)
地质过程与矿产资源国家重点实验室科技部专项经费资助
关键词
逐步证据权
加权证据权
LOGIT模型
云南个旧
stepwise weights of evidence
weighted weights of evidence
logit model
Gejiu of Yunnan Province.
