摘要
为了解绿泥石微量元素对不同成因矿床类型是否能够进行有效的分类判别,收集了13个来自斑岩型、矽卡岩型和浅成低温热液型3种类型矿床中的2 928条绿泥石微量元素数据,采用随机森林、支持向量机和人工神经网络3种不同的机器学习算法对矿床成因类型建立了分类模型并进行特征重要性分析.结果表明,依据Ni、Cr、Co、Sr、V、Zn 6种微量元素所建立的支持向量机模型分类效果最优,其Kappa系数最高为0.89,准确率、召回率和F1值的加权平均得分为0.96,Ni、V、Co为最关键的3个判别元素.基于绿泥石矿物微量元素结合机器学习方法能够实现对矿床类型的判别,对于区域尺度找矿勘查快速评价具有重要的指示意义.
Abstract
In order to study whether chlorite trace elements can effectively classify different genetic deposit types,in this paper,2 928 trace element data of chlorite from 13 different deposits were collected,which belong to three distinct genetic types,including porphyry,skarn and epithermal deposits.Three different machine learning algorithms,including random forest,support vector machine(SVM)and artificial neural network,were used to establish classification models for the genetic types of deposits and analyze the importance of characteristics.The results show that the SVM model based on Ni,Cr,Co,Sr,V and Zn,6 trace elements have the best classification effect,the highest Kappa coefficient is 0.89,the weighted average score of precision,recall and F1 value reach 0.96,and Ni,V and Co are the three most critical discriminant elements.In this paper it fully confirms that the machine learning classification model based on chlorite mineral trace elements can discriminate deposit types and provide an important indicator for the rapid evaluation of regional scale prospecting.
NETL
NSTL
万方数据