Prediction Model of Soil Electrical Conductivity Based on ELM Optimized by Bald Eagle Search Algorithm

Abstract

Soil electrical conductivity is one of the indispensable and important parameters in fine agriculture management, and a suitable soil electrical conductivity can promote good plant growth. Prediction model of soil electrical conductivity is constructed to effectively obtain the conductivity values of soil, which can provide a reference basis for irrigation and fertilization management and prediction evaluation in fine agriculture. Prediction model of soil electrical conductivity based on extreme learning machine (ELM) optimized by bald eagle search (BES) algorithm is proposed in this paper. In the prediction model, the input weights and bias values of the ELM network were optimized using the BES algorithm, and the performance of the model was evaluated with parameters such as mean square error (MSE), coefficient of determination (R^2). Also, the correlations of parameters such as soil temperature, moisture content, pH, and water potential in the soil conductivity prediction model were determined using the exploratory data analysis (EDA) and HeatMap heat map tools. Finally, the proposed model was compared with back propagation neural network (BP), radial basis function networks (RBF), support vector machine (SVM), gated recurrent neural network (GRNN), long short-term memory (LSTM), particle swarm algorithm (PSO) optimization ELM, genetic algorithm (GA) optimized ELM prediction model. The experimental results showed that MSE, R^2 of the proposed model are 4.09 and 0.941, which are better than the other models. Also the results verified the effectiveness of the proposed method, which is a feasible prediction method to guide the irrigation and fertilization management in fine agriculture, because of its good prediction effect on soil conductivity.