Forecasting the total power of China's agricultural machinery based on BP neural network combined forecast method

Abstract

In view of the limitations of single forecast model, forecasted results of different models will have some differences, in order to improve the forecast precision and the forecast results reliability, on the basis of determining the single forecast model for total power of China's agricultural machinery, the nonlinear combined forecast model for total power of agricultural machinery was established based on BP neural network using MATLAB software, and then the model was trained and simulated. The simulation results show that the fitting mean absolute percentage error of nonlinear combined forecast model is 0.59%, which is lower than 2.57%,2.66% and 2.09% of exponential model, GM(1,1) model and cubic exponential smoothing model .The established models were validated using original data of China's total power of agricultural machinery from 2009 to 2011, validation results show that the combined forecast model has the lowest forecast error 0.64%, the validation effect is the best, which can improve the forecast precision for total power of China's agricultural machinery. The total power of China's agricultural machinery was forecasted from 2012 to 2020 using the combined model, and the forecast results show that the total power of China's agricultural machinery will maintain a rapid growth trend in the next few years; it will be 1223987.1 MW by 2015 and 1603498.2 MW by 2020. © 2013 IFIP International Federation for Information Processing.