Prediction of Grain Yield Using SIGA-BP Neural Network

Abstract

In order to improve the accuracy of forecasting grain yield, detailed analysis of the reason that the BP network is vulnerable to fall into local minimum was made, then the new method was adopted to solve the problem of BP network. This paper studies the self-adaptive immune genetic algorithm (SIGA), and then uses the SIGA to optimize the BP neural network weights and thresholds values, used the SIGA global search method to solve the local minimum values of BP network, and meanwhile established the SIGA-BP network prediction model about Henan province’s grain yield. The simulation experiment results were that the average absolute error of grain yield predicted by the new model is 127.02ten thousand tons, the result shows that the SIGA-BP neural network model has higher prediction accuracy than the BP network model.