片烟库存预测研究中组合时间序列模型的应用

打开文本图片集

中图分类号：TP391.9 文献标识码：A DOI编码：10.3969/j.issn.1006—6500.2025.07.006

Abstract：Inordertoexplorethapplicationefectofhybridtimeseriesmodelsinpredictingtobaccostock，Monthlytobaostock dataofacertaincigareteinustryenterprisewerecolectedinthepast，singleandcombinationpredictionmodelsbasedontheexpo nentialsmoothingmethodandadaptivefilteringmethodwerebuilt，andtheefectsofdiferentmodelswerecompared.Theresults showedthatthemonthlyandannualtimeserieschartoftobaccostockdatashowedasiusoidaltrend withobvioustimeperiodicity. Thepredictionmodelbasedonte exponentialsmothing methodcouldnotjudgetherisingorfaling trendoftheealvalueaccurately. The RMSE，MAE and MAPE were 1.93，1.47 and 3.51% respectively. The prediction model based on the adaptive filtering method couldjudgethetrendofthepredictionobject，whilethepredictionerorwassignifcantlyraisedwiththeincreaseofsampledata.The RMSE，MAE and MAPE were 0.32，0.26 and 0.61% respectively.The predicted RMSE，MAE and MAPE of the combined time series model based on exponential smoothing method were 0.91，0.69， 1.75% and the predicted RMSE，MAE and MAPE of adaptive filtering combined time series model were 0.28，0.21， 0.52% respectively.In conclusion，the hybrid prediction model has better fitting effects， andcanbetereflcttheeallevelofflakytobaccostockdata.Amongthem，theadaptivefileringhybridtimeseriesmodelhsbetr prediction effect.

Keywords：flaky Tobacco;stock forecast;time series;hybrid model

预测片烟库存是烟草产业链先导环节，是一种受总量、结构和产地约束的战略资源。（剩余9668字）