Soil Heavy Metal Content Of Research On Intelligent Data Prediction Model

The enrichment of heavy metals in soils is of increasing concern to us.In order to prevent the threat of soil heavy metal pollution to public health,soil heavy metal pollution monitoring has become a very important task,and soil heavy metal content prediction is an important part of it.Machine learning is an important method for data prediction.The prediction error of a single machine learning model is often large,and the combined model of Machine Learning and Swarm Intelligence optimization algorithm is more conducive to data learning training.In order to better optimize the machine learning model,an improved Particle Swarm Optimization algorithm(C-PSO)is proposed in this paper,and the improved strategy aims to continuously generate more or better changing particles.First,to avoid the premature disappearance of particle diversity,C-PSO introduces randomly generated inertia weights between the interval [-1,1],and adjusts the calculation of the learning factor and the variation interval to accommodate the random inertia weight randomness.Then,a regenerable access strategy is proposed to allow populations that disappear during the merit search process to regenerate and enter the merit search space randomly.Finally,globally optimal particles are introduced so that the process of particle flight can better balance the effects of global and local optima on the particle search results,enabling the particles to fly toward better positions and yet avoiding premature particle maturation.In this paper,an Intelligent Data Prediction Model(IDPM)is proposed with Least Squares Support Sector Machine(LSSVM)as the basic model,while an improved Particle Swarm Optimization algorithm(C-PSO)is used to establish the iterative learning process of LSSVM for existing knowledge,and the learned model is used for prediction of new knowledge.In this paper,a small sample dataset of heavy metals from agricultural soils in East-West Lake and Hannan districts of Wuhan City is used as the experimental object,and the heavy metals Cr and Pb,which have a large impact on rice and vegetables,are selected for the content prediction experiments.The experiments were divided into three groups: comparison experiments of combined models of many PSO and LSSVM;combination experiments of many Swarm Intelligence algorithms and LSSVM;comparison experiments of three Neural Networks and SVM.Root Mean Square Error(RMSE),Mean Absolute Error(MAE)and Mean Absolute Percentage Error(MAPE)are used as evaluation indexes.The experimental results show that IDPM exhibits better prediction performance: for heavy metals Cr and Pb,the errors of IDPM’s MAPE in the East-West Lake area are 3.76% and 0.889%,respectively,and the errors of IDPM’s MAPE in the Hannan area are were 8.65% and 1.85%,respectively.