在BP神经网络算法的基础上,针对在多变量复杂系统建模过程中BP网络输入变量无法自动寻优的问题,将其与灰色关联分析方法结合,建立基于灰色关联分析的神经网络优化算法(GM-BPANN).并且使用中国粮食产量预测的数据,与逐步回归方法和灰色GM(1,N)模型方法进行了比较检验.结果表明,新模型通过灰色关联度的计算,可以全面、广泛地对大量的输入变量进行处理,而不必经过专门的主观筛选,从而增强了BP网络的适应能力,同时能够得到较好的预测精度和稳定性.
Aiming at the BP artificial neural network unable to auto select and optimize input variables, this paper integrates BPANN with grey relational analysis method, estabhshes an optimized BP artificial neural networkarithmetic (GM-BPANN) which based on the grey relational analysis method. And make comparison test with step regression method and grey GM(1,N) method using data of China grain output. The result shows that the new model can deal with mass input variables without special subjective selection, enhances the adaptability of BP neural network. It can also gets good forecasting stability and accuracy.