Prediction Model of Soil Heavy Metal Content Based on Particle Swarm Algorithm Optimized Neural Network

Area of Science:

• Environmental Science
• Geochemistry
• Computational Intelligence

Background:

• Soil heavy metal pollution in China exceeds 16.1%, necessitating focused control strategies.
• Ecological civilization initiatives prioritize soil heavy metal pollution prevention and control.

Purpose of the Study:

• To evaluate and compare the predictive performance of four neural network models for soil heavy metal content.
• To identify the most effective model for assessing soil heavy metal contamination in China.

Main Methods:

• Simulation and creation of four neural network models: Radial Basis Neural Network (RBFNN), Generalized Regression Neural Network (GRNN), Wavelet Neural Network (WNN), and Fuzzy Neural Network (FNN).
• Application of models to soil heavy metal data from two Chinese cities (northwest and central).
• Analysis of predicted vs. true values, prediction differences, and error indicators.

Main Results:

• The Wavelet Neural Network (WNN) demonstrated superior prediction accuracy compared to RBFNN, GRNN, and FNN.
• Performance evaluation involved comparing model predictions against actual soil heavy metal content data.
• Error indicator calculations confirmed WNN's enhanced predictive capabilities.

Conclusions:

• The Wavelet Neural Network (WNN) is the most effective model for predicting soil heavy metal content among the evaluated neural networks.
• Accurate prediction of soil heavy metal levels is crucial for effective environmental management and remediation efforts in China.