基于机器学习的场地重金属单因子指数预测  

作　　者：李心治 张健钦 胡昊 姜会忠 李星辰 陆楠 

机构地区：[1]北京建筑大学测绘与城市空间信息学院,北京102616 [2]自然资源部城市空间信息重点实验室,北京102616 [3]生态环境部信息中心,北京100029

出　　处：《环境保护与循环经济》2023年第2期35-38,共4页environmental protection and circular economy

基　　金：国家重点研发计划项目(2019YFC1804903);国家自然科学基金(41771413);北京市自然科学基金(8202013)。

摘　　要：为有效利用污染场地环境大数据,优化污染场地的调查与风险评估流程,依据污染行业专家知识建立指标集作为输入值,结合单因子指数污染评价方法,分别应用XGBoost,LightGBM两种机器学习模型对场地重金属单因子指数进行预测。模型训练和测试结果表明,两种模型均具有较好的预测效果,但整体上LightGBM略优于XGBoost。构建的场地重金属单因子指数预测模型可在无需钻孔采样分析的情况下,对场地内地块的重金属污染等级进行预判,为场地调查修复提供决策信息。In order to make effective use of the environmental big data of the contaminated site and optimize the investigation and risk assessment process of the contaminated site,the index set is established according to the pollution industry expert knowledge as the input value.Combined with the single factor index pollution assessment method,XGBoost and LightGBM machine learning models are respectively applied to predict the single factor index of heavy metals at the site.Model training and test results show that both models have good prediction effects,but LightGBM is slightly better than XGBoost on the whole.The single factor index prediction model of heavy metals in the site can predict the pollution level of heavy metals in the site without the need of borehole sampling analysis,and provide decision-making information for site investigation and restoration.

关 键 词：重金属污染 单因子指数法 机器学习 

分 类 号：X825[环境科学与工程—环境工程]