基于随机森林算法的川西木绒锂矿区及外围地电化学技术找矿预测  

Geoelectrochemical Prospecting Prediction in the Murong Lithium Deposit and Its Periphery in Western Sichuan Province Based on the Random Forest Algorithm

作　　者：甘盟 刘攀峰[1,2,3] 岳大斌 文美兰 高文[1,2,3] 邓鹏 陈若[1,2,3] 叶伟 GAN Meng;LIU Panfeng;YUE Dabin;WEN Meilan;GAO Wen;DENG Peng;CHEN Ruo;YE Wei(Institute of Concealed Deposit Prediction,Guilin University of Technology,Guilin,Guangxi 541006;Collaborative Innovation Center for Exploration of Nonferrous Metal Deposits and Efficient Utilization of Resources By the Province and Ministry,Guilin University of Technology,Guilin,Guangxi 541006;Guangxi Key Laboratory of Hidden Metallic ore Deposits,Guilin,Guangxi 541006;The 3rd Geological Brigade of Sichuan,Chengdu,Sichuan 610000)

机构地区：[1]桂林理工大学地球科学学院,广西桂林541006 [2]桂林理工大学,有色金属矿产勘查与资源高效利用省部共建协同创新中心,广西桂林541006 [3]广西隐伏金属矿产勘查重点实验室,广西桂林541006 [4]四川省第三地质大队,四川成都610000

出　　处：《地质与勘探》2025年第2期359-370,共12页Geology and Exploration

基　　金：深地国家科技重大专项(编号2024ZD1001503);国家自然科学基金(编号:42203067);广西高校中青年教师科研基础能力提升项目(编号:2023KY0258)联合资助。

摘　　要：随机森林算法具有能处理高维数据和缺失值,及对小训练样本集具有较高预测效果等优点,因而非常适合勘查地球化学相关的数据处理。本文以川西木绒锂矿区及外围为研究对象,利用地电化学技术采集到的地电化学数据,运用随机森林算法构建该区的找矿模型。将已知区地电化学技术采集到的Li元素及与其相关性强的Rb、Cs、Th元素和F1(Li-Rb-Cs-Th)组合元素作为训练指标对模型进行训练,得到本矿区的最佳随机森林模型,并对预测区的样本数据进行预测。经过多次对模型的训练,使已知区训练集和测试集的AUC值均大于80%,将模型运用于预测区,成功圈定了2处靶区。为检验靶区的准确性,对比Li、Rb、Cs、Th的单元素异常图及F1(Li-Rb-Cs-Th)组合元素异常图,所得到的综合异常区域与模型靶区位置一致,表明该随机森林算法预测模型具有较高的准确率,为川西木绒锂矿床的勘探工作提供了新的找矿方向。The random forest algorithm has the advantages of being able to handle high-dimensional data and missing values,and possesses a great prediction effect on small training sample sets,making it very suitable for data processing related to exploration geochemistry.This work took the Murong lithium deposit in western Sichuan and its periphery as the research object,and utilized the random forest algorithm to construct a prospecting model of this deposit.We selected the element Li,the strongly correlated elements Rb,Cs,Th,and the element combination F1(Li-Rb-Cs-Th)collected by the geoelectrochemical technology in the known area as training indicators to train the model.The best random forest model for this deposit was constructed,and the sample data in the prediction area were predicted.After multiple trainings of the model,the AUC values of the training set and the test set in the known area were both greater than 80%.The model was then applied to the prediction area,and successfully delineated two target areas.To verify the accuracy of the target areas,the single-element anomalies of Li,Rb,Cs,and Th and the element combination F1(Li-Rb-Cs-Th)were compared.The obtained comprehensive anomaly area is consistent with the location of the model target area,indicative of a high accuracy rate of the prediction model by the random forest algorithm.This study provides a new prospecting direction for the exploration work of the Murong lithium deposit in western Sichuan.

关 键 词：地电化学 随机森林模型 找矿预测 木绒锂矿 川西 

分 类 号：P62[天文地球—地质矿产勘探]