BP神经网络进行时间序列预测的不足及改进

BP算法是应用广泛的神经网络算法,具有较强的非线性拟合能力,可以用来预测非线性时间序列数据的发展趋势,在实际应用和仿真过程中,由于算法本身的限制和不足,对于仿真和计算都会带来这样那样的问题,比如网络训练过程中程序异常中止、训练时间过长、仿真精度不高等.针对这样的情况,通过分析算法本身和训练仿真过程,找到了相应的原因和解决方法,并通过完善训练过程,使问题得到了一定程度的改善.最后通过在Matlab仿真环境下的实际仿真过程,验证了改善效果.

Analysis and improving way of BP ANN in predicting time series data

BP(back propagation)algorithm is one of the most widely used neural network algorithm,and it has very high nonlinear fitting ability.So it can be used to predict the developing trend of time series data.In practical application and simulation,some kinds of problems and exception will happen because of the limitation and deficiency of the algorithm itself such as abnormal termination,long training time and low accuracy.Aiming at improving the performance,through analyzing the algorithm and simulation course,corre- sponding cause and problem-solving way is found.And through improving the simulating course,the performance of simulating is better. Finally,through practical simulating experiments in Matlab,the effect is certificated.