中文摘要：

当代矿产资源评价是复杂高维非线性系统的建模与评价过程,建立能够描述综合信息间复杂关系的多元非线性统计模型并预测矿产靶区,对矿产勘查具有重要指导意义.本文基于GIS软件平台,将证据加权模型应用于青海拉陵灶火地区矿产靶区预测,提取研究区成矿地质背景信息,遥感地质信息,地球化学异常信息等17种致矿综合信息,建立研究区矿产靶区预测模型,绘制成矿后验概率分布图,并利用ROC曲线分析方法对预测结果进行效果评价,根据ROC曲线的TP率和FP率计算圈定矿产靶区概率阈值,对研究区矿产靶区进行了圈定,结果表明证据加权和ROC曲线分析相结合圈定的成矿靶区与已知矿点分布较为吻合,该方法用于矿产靶区预测具有一定的可行性.

英文摘要：

The contemporary mineral resource evaluation is a process for the modeling and evaluation of the complex high dimensional non-linear system. Therefore, the multiple element non-linear statistic models established by describing the complex relationship of various kinds of comprehensive information to predict mineral target area have a great guiding significance in mineral resource exploration. Based on GIS software platform, the authors predicted mineral target area in Lalingzaohuo area of Qinghai Province by using weighted evidence model, and extracted 17 kinds of ore-generating comprehensive information such as metallogenic geological backgroud information, remote sensing geological information and geochemical abnormal information. The authors also established a model of mineral target area prediction in the study area, drew posterior probability distribution map, made an effect evaluation of the prediction result based on ROC curve analysis, calculated probability threshold for delineating mineral target area according to TP rate and FP rate of ROC curve, and delineated the mineral target area in the study area. The results show that the mineral target area delineated by the proposed method in combination with the weighted evidence and ROC curve analysis is quite in accord with the known ore spots, which suggests that this method has a certain feasibility for predicting the mineral target area.