Mineral Prospecting Information Mining And Mapping Mineral Prospectivity For Copper Polymetallic Mineralization In Southwest Fujian Province,China

The process of data mining and integrating for multi-source prospecting information played an important role in mineral resources exploration.It was considered as an important process for ensuring the effectiveness for mapping mineral prospectivity?MPM?.Therefore,it has always been the crucial research content to obtain the optimized prospecting information by geological data mining and integrating methods.For the past few years,the state-of-the-art machine learning algorithms were introduced into the field of MPM and has become the research hotspots and frontiers.Compared with traditional analytical methods,the machine learning approaches present wider range of applicability which benefit from the stronger capabilities for characterizing the relationship between variables and extracting the hidden distribution features.In the study,the southwest Fujian province of China was selected as a case study area and the multi-resource geological data are collected aiming to carry out the research of multi-resource prospecting information mining and MPM,which was based on GIS spatial analysis,fractal and multifractal theory and the machine learning methods.The main research contents and conclusions obtained in the study as follows:?1?The research of faults interpretation and controlling degree analysisThe geophysical data processing methods were used to interpret and improve the faults information on basis of aeromagnetic data and digital elevation model?DEM?data.Furthermore,the controlling degree of spatial distribution of copper deposits by different faults orientation was characterized by means of GIS-based spatial analysis and local singlarity analysis.It was demonstrated that the NW faults and NE faults show close spatial relationship with copper deposits;whereas the SN faults show negtive spatial relationship with copper deposits.?2?the selection of pathfinder elements and the identification of geochemical anomaliesThe following two studies have been carried out on basis of stream sediments geochemical data:?i?the influence of the secondary effects of geochemical elements was taking into account.The two aspects that the spatial distribution characteristics of geochemical patterns and the spatial relationship between geochemical elements and known mineral deposits have been considered.A hybrid method which combines the multifractal spectrum and the receiver operating characteristic?ROC?curve was applied to select the pathfinder elements;?ii?due to the decay and mask effects result from cover of vegetation,the spatial nonuniformity detected signals from the ground surface may be weak and this brings serious challenges to mineral exploration in these areas.Therefore,the deep autoencoder network was applied to identifying the concealed geochemical anomalies.It was demonstrated that the deep autoencoder network can be used for reducing the dimension of geochemical data and also was a powerful tool for recognizing the concealed geochemical anomalies.The result showed that the distribution of geochemical anomalies has close spatial relationship with known copper deposits.The method was helpful for improving the accuracy of geochemical anomalies identification.?3?The research of mapping intermediate and felsic intrusionsIt was difficult to obtain the completed geoinformation due to the decay and mask effects in the covered area.Therefore,the logistic regression model was applied to mapping intermediate and felsic intrusions on basis of the major elements and the aeromagnetic data in the study.The result demonstrated that the intermediate and felsic intrusions inferred by logistic regression model showed close relationship with outcropped intrusions and it can provid the new spatial information of intermediate intrusions for MPM.?4?The research of multi-source prospecting information mining and integrating based on machine learning methodsThe semi-supervised random forest method was applied to exploring the information of unlabeled data for solving the problem of the training samples?know mineral deposits?deficient.The case study demonstrated that it can make use of both the labeled and unlabeled samples to help improve the performance of MPM.To sum up,the main contributions of the dissertation as follows:?i?identification of geological anomalies associated with mineralization in the study area.The 20deep-seated faults were interpreted based on the aeromagnetic data and DEM data.Moreover,the major geochemical elements and the aeromagnetic data were integrated using logistic regression method to map intermediate and felsic intrusions in the study area.The research can provide the new geoinformation for MPM in the covered area;?ii?the multifractal spectral function and ROC curve were applied to selecting the pathfinder elements related to copper polymetallic mineralization.In terms of value of R,AUC and Z_AUC,the elemental association of Cu,Au,Ag,Pb,Zn,Fe₂O₃,W,Sn,Mo,Bi,P,MgO,Cd,and Sr can indicate occurrences of porphyry-epithermal copper mineralization in the study area.Furthermore,the deep autoencoder network was applied to identifying the concealed geochemical anomalies and it was considered to be an effective method for improving the accuracy of geochemical anomalies identification;?iii?the semi-supervised random forest model was applied to integrating the multi-source prospecting information for MPM.The four A-class,three B-class and two C-class metallogenic prospective areas were delineated and the results were helpful for guiding further mineral exploration for Cu polymetallic mineralization in southwest Fujian Province.