Prediction of Heavy Metal Content in Soil Based on AGA-Elman model

An AGA-Elman prediction model was proposed to solve the low prediction accuracy problem caused by data analysis outliers and the traditional Elman neural network in heavy metal content prediction. The outliers were detected and eliminated by the method based on entropy weight distance to reduce the influence of outliers on the fitting of high-dimensional data models. To solve the defects of the original Elman network model, such as easily falling into the local minimum when predicting y, a Genetic Algorithm (GA) was used to improve it, and adaptive laws were introduced to improve the global optimization ability of the algorithm. The mean absolute error and mean absolute percentage error of the AGA-Elman model in predicting the heavy metal content of AS are 1.811 and 20.05%, respectively, and the determination coefficient index is 5.1% higher than that of the comparison model. The experimental verification of several different prediction models shows that AGA-Elman has higher accuracy in predicting soil heavy metal content, which provides a more accurate and faster method for predicting soil heavy metal content.