Related Experiment Video
Updated: Aug 9, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.9K
Improving the efficiency of machine learning in simulating sedimentary heavy metal contamination by coupling
Ligang Deng1, Xiang Gao2, Bisheng Xia3
1State Key Laboratory of Pollution Control and Resource Reuse, School of Environment, Nanjing University, Nanjing, 210023, China.
Chemosphere
|February 23, 2023
Summary
Heavy metal pollution in Taihu Lake surged after the 1970s due to human activities. Machine learning models effectively predicted heavy metals using sediment magnetic properties, aiding environmental monitoring.
Area of Science:
- Environmental Science
- Geochemistry
- Data Science
Background:
- Taihu Lake, a significant freshwater body in China, faces environmental challenges.
- Sediment cores provide a historical record of environmental changes and pollution events.
- Understanding heavy metal sources and developing predictive models are crucial for lake management.
Purpose of the Study:
- To investigate the historical trends of heavy metal pollution in Taihu Lake sediments.
- To explore the relationship between heavy metal concentrations and sediment magnetic properties.
- To develop and evaluate machine learning models for predicting heavy metal levels using selected sediment parameters.
Main Methods:
- Sediment cores were collected and dated using radionuclide analysis.
- Heavy metal concentrations and magnetic properties of sediment samples were measured.
- Feature selection methods (Random Forest and Maximal Information Coefficient) were combined with Support Vector Machine (SVM) for predictive modeling.
Main Results:
- A significant increase in heavy metal concentrations was observed around the 20 cm depth, coinciding with rising single-domain magnetic particle concentrations, suggesting anthropogenic influence post-1970s.
- Machine learning models using selected magnetic and physicochemical parameters achieved reasonable simulation performance for heavy metals.
- Random Forest (RF) outperformed Maximal Information Coefficient (MIC) in improving simulation accuracy (R² values) for several heavy metals (Cd, Cr, Cu, Pb, Sb).
- High correlation coefficients (0.73–0.97) were achieved for predicting ecologically risky heavy metals (As, Cd, Cr, Hg, Pb, Sb) using only 14–27% of parameters selected by RF.
Conclusions:
- Anthropogenic activities since the 1970s have significantly impacted Taihu Lake's heavy metal pollution levels.
- The integration of feature selection methods with machine learning, particularly RF-RBF-SVM, provides a robust approach for predicting heavy metal concentrations in lake sediments.
- Sediment magnetic properties are valuable indicators for predicting heavy metal pollution, offering a cost-effective and efficient monitoring tool for lake ecosystems.
