摘要
针对订单数据的非线性特性,以及受市场动态波动影响,且常常会存在训练样本有限的情况,提出采用ANN定性预测法,即神经网络与定性预测结合的方法预测订单。在传统BP神经网络的基础上引入牛顿法与竞争学习算法,以提高收敛速度并改善传统BP神经网络容易陷入局部极小值的情况。根据订单预测的实际情况列出三种定性预测与神经网络的结合形式,以及各形式的适用情况,最后通过实例分析表明该方法可行,并能有效地提高订单预测的精度。
According to the data of order’ s nonlinear characteristics, as well as be affected by the dynamic market and exist the situation of training samples limited , the method of ANN qualitative forecasting is proposed that the neural network combined with the qualitative forecasting method to forecast orders. This paper introduced the Newton method and competitive learning algorithm based on traditional BP neural network to improve the convergence rate and the circumstance of traditional BP neural network is easy to fall into local minimum values. With the actual situation of order forecast, list three forms of qualitative forecast combined with neural networks and the application of them. Finally, through the analysis of example show the method is feasible and can effectively improve the forecasting accuracy.
出处
《机电工程技术》
2014年第9期23-26,95,共5页
Mechanical & Electrical Engineering Technology
关键词
BP神经网络
定性预测
订单预测
结合
BP neural network
qualitative prediction
order forecasting
combination