高斯核密度估计连续后验Logit模型找矿方法  

Gaussian kernel density estimation continuous post-testing Logit model prospecting method

作　　者：浦路平[1] 尹意求 肖克炎[3] 汪新庆[4] 林锦富 梁世华 PU Luping;YIN Yiqiu;XIAO Keyan;WANG Xingqing;LIN Jinfu;LIANG Shihua(School of Information Science and engineering,Guilin University of Technology,Guilin 541006,Guangxi,China;China Nonferrous Metals(Guilin)Geology and Mining Co.,Ltd.,Guilin 541004,Guangxi,China;Institute of Mineral Resources,Chinese Academy of Geological Sciences,Beijing 100037,China;Faculty of Earth Resource,China University of Geosciences,Wuhan 430074,Hubei,China)

机构地区：[1]桂林理工大学信息科学与工程学院,广西桂林541006 [2]中国有色桂林矿产地质研究院有限公司,广西桂林541004 [3]中国地质科学院,北京100037 [4]中国地质大学(武汉)资源学院,湖北武汉430074

出　　处：《矿产与地质》2023年第4期866-874,共9页Mineral Resources and Geology

基　　金：国家自然科学基金项目(41562018);桂林理工大学博士基金共同资助。

摘　　要：针对证据权模型存在连续型数据需要划分为二态数据并要检验自变量之间的条件独立性的问题,提出了连续后验Logit二分类模型,并据此提出了高斯核密度估计连续后验Logit模型找矿方法,该方法用连续型数据直接计算有矿后验Logit,无需转换连续型数据为二态数据,也无需检验变量之间条件独立性,减少了工作步骤,可以提高定量找矿的精度和工作自动化程度。利用化探数据对广西丹池成矿带的锡矿预测与已知矿产空间分布具有较好的对应关系,证明该方法是可行和有效的。In order to solve the problem that continuous data needs to be divided into two state data and the conditional independence between independent variables exists in the evidence weight model,this paper proposes a continuous posteriori Logit classification model and prospecting method based on Gaussian kernel density estimation.This method uses continuous data to directly calculate the ore-bearing post-testing Logit,without converting continuous data into binary data,and without testing the conditional independence between variables,which reduces the work steps,and can improve the accuracy and automation of quantitative prospecting.It is proved that the method is feasible and effective by using geochemical data to predict tin deposits in Danchi metallogenic belt,Guangxi,which has a good correspondence with the spatial distribution of known mineral lands.

关 键 词：连续后验Logit分类模型 高斯核密度估计 证据权模型 

分 类 号：P612[天文地球—矿床学] TP319[天文地球—地质学]