基于改进BP神经网络的饮料销售预测模型研究

摘要

In order to accurately predict the demand of beverage sales, aprediction model of beverage sales based on Grey rough set and BP neural network was established.Aiming at the non-linearity, redundancy and incompleteness of neural network input data, the grey relational analysis and rough set reduction are used to deal with the two-dimensional data, which can improve the training speed and generalization ability of BP neural network.Aiming at the slow learning convergence speed and local minimum error of BP neural network, the learning convergence speed and prediction precision of BP neural network are improved by adding momentum term and optimizing error function.An example is given to verify the model with the historical sales data of a beverage enterprise.The results show that the model has smaller error and higher prediction accuracy.%为准确预测饮料销售量, 构建了一个灰色粗糙集与BP神经网络结合的饮料销售预测模型.针对神经网络输入数据的非线性、冗余性、不完整性, 通过灰色关联分析和粗糙集属性约简对其进行双维度处理, 提高了BP神经网络的训练速度和泛化能力.通过附加动量项、优化误差函数, 提高了BP神经网络的学习收敛速度和预测精度.针对某饮料企业历史销售数据进行了实例验证.结果表明, 该模型相比未经前期处理的BP神经网络模型和线性回归分析方法, 其预测结果的误差更小, 预测精度更高.