A Combined Forecasting Method of Grain Yield in China Based on GM(1,1) and BP Network

In view of the fact that the system of grain yield is affected by many factors and has complicated non-linear characteristic, a combined forecasting model by using multi-indicator for grain yield in China is constructed based on BP network and grey system, which can be named GM(1, 1)-BP model. Seven index were chosen from agricultural production conditions, he primitive data of the multi-factors from 1980 to 2006 are taken as the input of BP network, The primitive data of the grain yield from 1980 to 2006 are taken as the output. Then the network structure, initial weighted values and thresholds are set. Taking the forecasting results of GM(1, 1) models for every factor from 2007 to 2015 as the input of BP network, the corresponding output of simulation are the forecasting results of the GM(1, 1)-BP model, that is the grain yield from 2007 to 2015. The data from 2007 to 2008 are used as test sets, empirical results show that the combined model has higher precision and training efficiency than the models based on GM(1, 1), BP network or GM(1, N) alone.