改进BP算法在预测安全库存中的应用

Improved Algorithm of BP Neural Networks on Predicting Safety Stock

针对标准BP算法收敛速度慢、容易陷入极小值的缺陷,利用附加动量项和变步长思想相结合的改进算法,建立三层BP网络模型对安全库存量进行预测,并与标准BP算法的预测结果进行比较,提高了网络收敛速度和预测准确度。

To solve the drawbacks of BP algorithm such as slow convergence and often going to error status of local minima,this paper combines idea of additional momentum item and variable step to building three-layer BP network model to predict safety stock.Compared to the standard BP algorithm,the improved BP algorithm greatly improves the network convergence rate and prediction accuracy.