基于机器学习的土壤蔬菜重金属迁移模型研究  

Research on Migration Model of Heavy Metals in Soil Vegetable Systems Based on Machine Learning

作　　者：姜传智 刘翠玲[1,2] 孙晓荣[1,2] 张善哲 JIANG Chuan-zhi;IU Cui-ling;SUN Xiao-rong;ZHANG Shan-Zhe(College of Artificial Intelligence,Beijing Technology and Business University,Beijing 100048,China;Beijing Key Laboratory of Big Data Technology for Food Safety,Beijing Technology and Business University,Beijing 100048,China)

机构地区：[1]北京工商大学人工智能学院,北京100048 [2]北京工商大学食品安全大数据技术北京市重点实验室,北京100048

出　　处：《计算机仿真》2025年第1期422-429,共8页Computer Simulation

基　　金：北京市自然科学基金项目(4222043)。

摘　　要：随着工业的飞速发展,土壤受重金属污染的压力越来越大,进而对农作物的种植造成了严重威胁。通过机器学习算法建立土壤-蔬菜生态系统重金属迁移预测模型,结合特征重要性分析评价了蔬菜生产系统生态健康风险。研究中共收集160组样本数据,分别采用支持向量回归(SVR)、极限学习机(ELM)、径向基函数神经网络(RBF)以及随机森林(RF)算法构建不同的土壤-蔬菜生态系统重金属迁移预测模型,并通过特征重要性分析研究了土壤理化因子对蔬菜重金属迁移的影响。与其它3种机器学习算法所建立的模型相比,基于RF算法建立的蔬菜重金属迁移预测模型的R^(2)为0.92,RMSEP为0.0093。研究结果表明,土壤重金属有效态含量(SCB)和土壤重金属总含量(SC)是影响蔬菜重金属富集的主要因素。研究k中采用机器学习算法验证了利用迁移模型预测蔬菜重金属含量的可行性,为构建蔬菜产地重金属污染预防技术体系提供理论支撑。With the rapid development of industry,heavy metal pollution has caused significant damage to the soil,which resulted in a serious threat to crop cultivation.A heavy metal migration prediction model for the soil vegetable ecosystem was established by machine learning algorithms.The ecological health risks of the vegetable production system were also evaluated through feature importance analysis.This study collected 160 sets of sample data.In this work,different machine learning algorithms were used to set various prediction models for heavy metal migration in soil vegetable ecosystems,such as support vector regression(SVR),extreme learning machine(ELM),radial basis function neural network(RBF),and random forest(RF)algorithms.The influence of soil physicochemical factors on heavy metal migration in vegetables was also studied through feature importance analysis.Compared to the other models,the R^(2) and RMSEP of the model based on the RF algorithm both obtained the optimal solutions,which were O.92 and 0.0093,respectively.The experimental results indicated that the available heavy metal content(SCB)and total heavy metal content(SC)in soil were the main factors for the accumulation of heavy metals in vegetables.Furthermore,transfer models were used to predict the heavy metal content of vegetables.The feasibility of machine learning algorithms was also verified in vegetable risk assessment.This work provided theoretical support for constructing a technical system for preventing heavy metal pollution in vegetable-producing areas.

关 键 词：重金属污染 土壤-蔬菜生态系统 重金属迁移 机器学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]