机器学习在矿物岩石地球化学大数据挖掘中的应用与展望

打开文本图片集

Abstract： With the rapid advancement of information technology，geoscience research has entered the data-intensive era， where the large-scale mining of mineraland rock geochemical data has become a critical approach for deciphering geological evolution， deepening our understanding of ore-forming processes and enhancing exploration efficiency. As arepresentative data-driven modeling technique， machine learning ofers powerful capabilities for identifying hiden paterns and key features within complex datasets， thereby providing new methodological avenues for geoscientific studies.This paper systematically reviews the general workflow of mineral-rock geochemical big data mining; outlines the basic principles of representative machine learning algorithms; and evaluates their reent applications in tectonic seting discrimination，rock genesis and evolutionaryreconstruction， mineral prospectivity assessment，and ore deposit type classification. Onthis basis，we have summarized the strengths and limitations of machine learning in geochemical data mining，highlighting challnges such as limited sample sizes，uneven data distribution，and insufcient model interpretability.Furthermore，wediscussfutureresearch prospects inalgorithm optimization，the development of mineral-rock geochemical databases，the integration of deep learning and transfer learning， multisourcedata fusion，interpretableartificial inteligence，and theapplicationof low-codeframeworks.The widespread application of machine learning in mineraland rock geochemical data mining is expected to provide robust theoretical supportand methodological pathwaysforthequantitativeanalysisof geological processs，inteligent decision-making in mineral exploration，and the systematic development of Earth science research.

Keywords：machine learning；mineral and rock geochemistry；big data mining；mineral prospectivity assessment；ore deposit type classification

21世纪以来，人工智能、机器学习和深度学习等技术迅速发展，广泛应用于计算机视觉、自然语言处理等多个领域（Ngai etal.，2011;Esteva etal.， 2019; Fan et al.， 2020; Haydari and Yilmaz， 2022;成秋明，2022），显著提升了数据处理效率与人类活动的智能化水平。（剩余31911字）