改进的BP神经网络及其在销量预测中的应用

Sales forecasting model based on improved BP neural network

针对影响产品销量的因素众多,并且影响因素之间相互作用等特点,将人工神经网络理论引入产品销量预测领域.同时,为了克服BP网络的局限性,提出将主成分分析方法、BP神经网络以及粒子群优化算法相结合,分别从样本质量和初始权值两个方面对BP神经网络进行改进.最后,对某品牌服装产品的月销售量进行了实例研究.结果表明,所提出的模型简化了BP网络结构的同时,提高了网络的预测精度,从而验证了模型的有效性.

In the light of different factors affecting sales of products and the interaction between those factors,the theory of artificial neural network was introduced into the domain of sales forecasting.At the same time,BP neural netowrk was improved both from the aspects of sample data quality and initial parameters to overcome its limitation by combining principal components analysis（PCA）,BP neural network and particle swarm optimization algorithm（PSO）.Finally,an example analysis was made in order to verify the validation of this model.The results showed that the suggested model simplified the architecture of BP network and improved forecast accuracy.Thereby,the effectiveness of this model was validated.