基于Levenberg-Marquardt算法改进BP神经网络的卷烟销量预测模型研究

摘要

A BP neural network improved by Levenberg-Marquardt algorithm was illustrated, established and applied to forecast cigarette sales to overcomie disadvantages of general time series analysis. Cigarette sales data were normalized and repeated training and simulating on the model with Matlab software. Compared with real sales data, the forecast of the improved BP neural network is proved accurate.%针对一般时间序列分析方法中预测方法的不足,采用改进的BP神经网络对卷烟销量进行预测.介绍说明改进的BP神经网络Levenberg- Marquardt算法原理,对卷烟销量数据进行归一化处理,建立卷烟销量神经网络预测模型,利用Matlab软件对数据进行训练、仿真.与实际销量进行对比分析,证明采用改进的BP神经网络预测结果准确.