Mineral search prediction based on Random Forest algorithm ——A case study on porphyry-epithermal copper polymetallic deposits in the western Gangdise meatallogenic belt
OUYANG Yuan,LIU Hong,LI Guangming,MA Dongfang,ZHANG Linkui,HUANG Hanxiao,ZHANG Jinghua,ZHANG Tengjiao,LIU Xiao,ZHAO Yinbing,LI Fu
DOI: https://doi.org/10.12029/gc20201026001
2023-01-01
Abstract:The paper is the result of geological survey engineering.[Objective] The core problem of prospecting prediction is the nonlinear relationship between mineral distribution and mineralcontrolling geological factors. Big data and machine learning technology have shown great advantages in solving such complex nonlinear relationship problems. The prediction dataset of small-scale geochemical remote information has the characteristics of high and extremely unbalanced, which is difficult to adapt by traditional logical assumptions or statistical analysis. Therefore, this paper attempts to introduce the random forest algorithm into the field of small-scale prospecting to explore the application of big data and machine learning technology in small-scale mineralization prediction. [Methods] In recent years, several Porphyryepithermal copper polymetallic deposits（such as Luerma, Bolazha, Daruo, Hongshan, and Luobuzhen, etc.） have been discovered in the western Gangdise mineralized belt, which proved that the western Gangdise belt has great prospecting potential for porphyry and epithermal Cu-Au polymetallic deposits. Combined with the comprehensive information of typical deposits, regional geology,geophysics, geochemistry, and remote sensing, this paper uses the random forest method to carry out the prospecting prediction of porphyry and epithermal Cu-Au polymetallic deposits in the western Gangdise belt. [Results] This work has delineated 11 porphyry copper polymetallic prospect areas（including 2 levels I prospect areas, 3 level II prospect areas, and 6 level III prospect areas）, of which Luobuzhen, Dajiacuo, Daruo, Balaza, Gaerqiong, and Budongla have great prospecting potential and are expected to find new ore deposits or points. [Conclusions] The under-sampling random forest prediction model based on big data machine learning is expected to adapt to the high-dimensional and extremely unbalanced characteristics of prediction data of comprehensive geophysical and geochemical remote information and provide direction for regional prospecting prediction at the scale of the metallogenic belt. The prospective area determined in this work is expected to find new deposits（points）, which opens a new vision for ore prospecting and exploration in the Gangdise metallogenic belt.