The Statistical Properties And Application Of Judgmental Hierarchical Forecasting Model

Vending machine sales forecast is a small part of the construction of smart city,the city is now planning to establish one used to supplement more stores inventory of the distribution center,the distribution center to the jurisdiction of all vending machine regularly provide goods,then each vending machine need how many goods,how many goods distribution center need to be added,in different commodity scenarios to choose what kind of forecast method,which is the problem I want to study.Obviously store operations require store-level forecasting,but inventory decisions in distribution centers require overall forecasting.We adopt the exponential smoothing prediction method,which has two models:the top level prediction model and the bottom up prediction model.When the smoothing coefficient is optimal,the two models are equivalent,so we add the random judgment error term into the model.The content of this paper mainly includes the following three points:I.Based on the combination of hierarchical prediction and judgment prediction,this paper explores which hierarchical prediction model is more advantageous in different sales scenarios.2.We generated artificial data.These data first met the basic conditions of the hypothesis,and grouped them into groups to make each group correspond to a sales scenario.3.At last,we grouped the real vending machine sales data to verify our hypothesis.Inspired by the literature I read,we compared the advantages and disadvantages of ARMA model and exponential smoothing model in different scenarios and presented the results at the end of the fourth chapter.At the end of the paper,we discuss and summarize.The research data in this paper comes from real sales records.We use data analysis method to convert the original data into data suitable for prediction,and then use data mining method to conduct experimental analysis on python software.