基于贝叶斯模型与VAR模型的蔬菜销量与定价预测

Vegetable Sales and Pricing Forecasting Based on Bayesian Model and VAR Model

对商超蔬菜日销售数据进行数据透视筛选,利用同期平均法处理异常值和缺失值,生成年度日销售新数据.对单品蔬菜销量进行Shapiro-Wilks检验,发现不服从正态分布,但经相关性分析发现同类别不同蔬菜单品销量间存在显著相关性.构建同类蔬菜销量Bayesian模型,定量化表达单品间销量的关系,建立单品蔬菜销量的Bayesian模型刻画销量的主要影响因素,两个模型分析和预测结果与实际相符.构建VAR(2)模型,完成未来一周补货设计.以收益最大化为目标,在满足市场需求与销售空间限制下,建立优化模型,完成未来一天的补货量及定价设计,为商超蔬菜销售提供合理化建议.

The daily sales data of supermarket vegetables were screened through data perspective,and the outliers and missing values were processed by the same period average method to generate new annual daily sales data.Shapiro-Wilks test was conducted on the sales volume of single vegetable,and it was found that the distribution did not follow normal distribution,but the correlation analysis found that there was a significant correlation between the sales volume of different vegetable items of the same category.This paper constructs a Bayesian model of similar vegetable sales,quantifies the relationship between single product sales,and establishes a Bayesian model of single product vegetable sales to describe the main influencing factors of sales.The analysis and prediction results of the two models are consistent with the reality.Build the VAR(2)model and complete the replenishment design in the coming week.With the goal of maximizing revenue,under the condition of meeting market demand and sales space limitation,the optimization model is established to complete the replenishment volume and pricing design of the next day,and reasonable suggestions are provided for the vegetable sales of supermarket.