Research On Prediction Method Of Soil Heavy Metal Content Based On Neural Network And Intelligence Algorithm

When controlling soil heavy metal pollution,researchers often need to master detailed soil heavy metal pollution information.However,due to the limitations of manpower and material resources,it is difficult to obtain heavy metal detection results in all areas.Therefore,this study mainly uses local area data sampling information to predict the soil heavy metal content in unknown areas.The existing prediction methods of soil heavy metal content can obtain better prediction results when dealing with datasets with more eigenvalues,and once the eigenvalues of the datasets are reduced,the prediction effect will decrease.Therefore,this paper carries out related research based on neural network and swarm intelligence optimization technology,and at the same time a deep composite neural network model is proposed for soil heavy metal content prediction,aiming to improve the prediction accuracy of soil heavy metal content when the dataset has fewer eigenvalues,and to provide relevant basis for the government when formulating soil heavy metal pollution control policies.The research content of this article mainly includes the following aspects:(1)Firstly,four kinds of artificial neural networks: radial basis function neural network(RBFNN),generalized regression neural network(GRNN),wavelet neural network(WNN)and fuzzy neural network(FNN),which are commonly used in data prediction,are selected to build prediction models.Then,in the same experimental environment,compare the prediction performance of these four models on two sets of soil heavy metal content data after processing,and finally select the best neural network model based on the comparison results.(2)Aiming at the problem that the initial parameters of the artificial neural network model used for data prediction are difficult to determine,the relevant research on the swarm intelligence optimization algorithm for parameter optimization is carried out.Firstly,considering factors such as the order of the proposed years,the commonness and algorithm optimization mode,four swarm intelligence optimization algorithms are selected: genetic algorithm(GA),firefly algorithm(FA),grey wolf optimizer(GWO),and bird swarm algorithm(BSA).Then,on the basis of these four algorithms,the problems of each algorithm are improved in a targeted manner,and new algorithms are proposed accordingly.At the same time,the optimization results of the algorithms on a variety of different test functions are compared in the same experimental environment.Finally,the optimal swarm intelligence optimization algorithm is selected through the comparing results.(3)Combine the best-performing artificial neural network with the best-performing swarm intelligence optimization algorithm,and at the same time improve the back-propagation method of the artificial neural network and propose an adaptive root mean square back-propagation(ARMSProp),which is combined into deep composite neural network model(DCNN).Finally,in the same experimental environment,this model was compared with several common soil heavy metal content prediction models: ordinary least squares regression model(OLS),support vector machine model(SVM),Bayesian model,and the best-performing artificial neural network model in two different soil heavy metal datasets to verify the effectiveness of the prediction method proposed in this paper.