Research And Application Of Regional Soil Heavy Metal Pollution Prediction Model

Soil heavy metal contamination threatens human health.However,traditional soil heavy metal contamination detection methods are costly,low in efficiency.Besides the problem of missing elemental data of collected soil heavy metal samples is serious,the missing value supplementation methods in the field of ecology and environment have low accuracy and high limitation.To solve above problems,a regional soil heavy metal contamination prediction model is built.This thesis takes a certain region in the North China Plain as the research region,collects the data of 8 types of soil heavy metals in the study region in 2020,uses these 8 types of soil heavy metal data to make a comprehensive and systematic assessment of the study region.The main research work is as follows:1.A method of using machine learning to interpolate the missing values of sample data is proposed,and a soil heavy metal missing value prediction model based on the improved BP neural network is constructed to predict the missing soil heavy metal elements Cd,Hg,and Pb.In order to improve the shortcomings of the traditional BP neural network,the Fruit Fly Optimization Algorithm is optimized by using the Simulated Annealing algorithm,and the optimized SA-FOA algorithm is used to replace the parameter optimization method in the traditional BP neural network.Compared with the traditional BP neural network algorithm,the prediction error of SA-FOA-BP neural network model proposed in this chapter reduces by nearly 10%.2.A method of using small sample data to predict regional soil heavy metal contamination is proposed,and a soil heavy metal contamination prediction model based on RF-GA-SVR is constructed.The Random Forest algorithm is used to extract and train the characteristics of the heavy metal content in the soil to obtain the classification results of the soil heavy metal contamination status,the results of Random Forest feature selection optimization are then input into the Support Vector Regression model,and then the parameters of the Support Vector Regression model are optimized using a Genetic Algorithm,which leads to the prediction of soil heavy metal contamination.3.A contamination source classification model based on an improved Self-Organizing feature Mapping neural network is constructed.Based on the soil heavy metal data of the study region,combined with relevant literature and topographic data of the study region,correlation analysis and Principal Component Analysis are used to evaluate and classify the ecological hazard coefficients of 8 heavy metal elements in soil samples.The above heavy metals are analyzed by improving the PCA-SOM algorithm for traceability.The regional soil heavy metal contamination prediction model is constructed by soil heavy metal missing value prediction method,soil heavy metal contamination prediction method and soil heavy metal pollution source classification method.In summary,the study of regional soil heavy metal contamination prediction model has high research significance and practical value.