Scholarlink

Purpose <p>The traditional method to conduct the soil pollution survey usually needs massive sampling and testing which is very costly. New technology with low cost is needed.</p> Methods <p>The data-driven machine learning method was employed to predict the heavy metal concentration in the present study.</p> Results <p>In the present work, machine learning methods were used to predict the heavy metal concentration in soil from sites in Zhejiang Province which locates on the southeast coast of China. Six machine learning models were employed, including Linear Regression (LR), Support Vector Machine (SVM), Random Forest (RF), eXtreme Gradient Boosting (XGB), Backpropagation Neural Network (BP) and Radial Basis Function Neural Network (RBF). The evaluation of the machine learning prediction shows that the best performances are from the Bayesian-optimized RF(RF_BO) and XGB models. Moreover, the stacking model based on the RF_BO and XGB models achieved high accuracy (R² &gt;0.8) for Cr, Cu, Ni, Pb, and Zn.</p> Conclusion <p>Machine learning methods can make good predictions about contamination in the studied sites, which can significantly improve the efficiency of heavy metal contamination testing and reduce the cost.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning prediction of the heavy metal concentration in soil

  • 期刊论文

  • 发表日期：

    2025年9月24日

查看全文
  • Jinfeng Chen,
  • Wei Zhang,
  • Yue Peng,
  • Junliang Wang,
  • Min Wu

摘要

Purpose

The traditional method to conduct the soil pollution survey usually needs massive sampling and testing which is very costly. New technology with low cost is needed.

Methods

The data-driven machine learning method was employed to predict the heavy metal concentration in the present study.

Results

In the present work, machine learning methods were used to predict the heavy metal concentration in soil from sites in Zhejiang Province which locates on the southeast coast of China. Six machine learning models were employed, including Linear Regression (LR), Support Vector Machine (SVM), Random Forest (RF), eXtreme Gradient Boosting (XGB), Backpropagation Neural Network (BP) and Radial Basis Function Neural Network (RBF). The evaluation of the machine learning prediction shows that the best performances are from the Bayesian-optimized RF(RF_BO) and XGB models. Moreover, the stacking model based on the RF_BO and XGB models achieved high accuracy (R² >0.8) for Cr, Cu, Ni, Pb, and Zn.

Conclusion

Machine learning methods can make good predictions about contamination in the studied sites, which can significantly improve the efficiency of heavy metal contamination testing and reduce the cost.