城市地质2025，Vol.20Issue(3)：301-309,9.DOI:10.3969/j.issn.2097-3764.2025.03.003
有监督与无监督式机器学习在含水层脆弱性评价中的应用
Application of supervised and unsupervised machine learning in aquifer vulnerability assessment
摘要
Abstract
This study conducts a comparative multi-method aquifer vulnerability assessment for the Mi-Huai-Shun area.We establish three distinct models:the traditional DRASTIC model,an unsupervised K-means clustering model,and a supervised XGBoost machine learning model.Findings reveal that the aquifer vulnerability in the study area is dominantly characterized by high,moderately high,and medium levels.The DRASTIC and K-means models show strong consistency,identifying high-vulnerability zones primarily in the central-northern regions and along riverbanks,thereby confirming hydrogeological conditions as a key control on vulnerability zoning.In contrast,the XGBoost model identifies a more extensive distribution of high-vulnerability zones.Leveraging its machine learning capabilities,XGBoost effectively captures the nonlinear influences of key feature indicators,including soil type,aquifer media structure,and groundwater depth.Methodological comparisons demonstrate that the DRASTIC model facilitates rapid assessment via its clear indicator system,K-means achieves objective zoning through data self-organization,and XGBoost excels in characterizing complex geological processes.关键词
地下水/脆弱性评价/机器学习/密怀顺/应用比较Key words
groundwater/vulnerability assessment/machine learning/Mi-Huai-Shun area/application comparison引用本文复制引用
陈凯,杨庆,徐庆勇,赵微,张旭航..有监督与无监督式机器学习在含水层脆弱性评价中的应用[J].城市地质,2025,20(3):301-309,9.基金项目
北京市地下水监测网运行项目(11000022T000000440194)资助 （11000022T000000440194）
