Prediction Of Heavy Metal Content In Soil Based On Adaptive Evolution Model

In the prevention and control of soil pollution by heavy metals,it is difficult to obtain information about the content of heavy metals in the soil in all fields of study.Therefore,in this study,we used a method to predict soil heavy metal content over a wide area with a small amount of information from the sample data.In this paper,we propose an adaptive evolutionary model(AEM)that predicts heavy metal content in soil based on neural networks and intelligent algorithms,and improves the accuracy of predicting heavy metal content in soil with the aim of solving the problem of deterioration in prediction accuracy due to reduced characteristics in the set data.The research content of this article is as follows:(1)Use the Elman waveform neural network,short-term memory neural network and bidirectional neural network to create a prediction model,compare and analyze the prediction results of the four models,and select the neural network with the best prediction performance.(2)To improve the accuracy of the traditional neural network model when predicting heavy metal content in soil,the parameters of the Q network,wavelet neural network and recursive neural network parameters were optimized.Various methods improve three types of networks and establish three prediction benchmarks,namely: a distance-weighted self-adaptive inverse interpolation model,a wavelet neural network model based on a common algorithm of bird flocks,and a bidirectionally closed neural network model based on the mechanism of self-attention.(3)Based on the powerful Elman neural network,it is the best predictive experiment above,using the gray wolf algorithm to improve performance and provide a model of adaptive evolution.This model first uses Bayes regularization to improve the objective function of the Elman network model,Then uses the Gray-Wolf adaptive algorithm to optimize the default parameters of the network model.Entropy weights,which detect irregular values,are used to eliminate anomalous values in the data.Based on the heavy metal content data collected by the domestic agricultural sector in question,the average absolute error of the adaptive evolutionary model proposed in this article is 1,623,and the average absolute percentage error is 17.48%,0.394% higher than Elman’s model in determining the coefficient index.Experimental validation of five different prediction models showed that adaptive evolutionary models are more accurate in predicting heavy metal content in soil,providing a more accurate and efficient prediction method for heavy metal contamination on agricultural land.