基于GWO-Prophet的商品销售预测研究  

Research of Commodity Sales Prediction Based on GWO-Prophet

作　　者：曾文烜 高永平[1] ZENG Wenxuan;GAO Yongping(College of Information Engineering,East China University of Technology,Nanchang 330013)

机构地区：[1]东华理工大学信息工程学院,南昌330013

出　　处：《计算机与数字工程》2024年第3期659-664,699,共7页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:11865002);江西省教育厅科学技术研究项目(编号:104506);东华理工大学江西省放射性地学大数据技术工程实验室项目(编号:JELRGBDT201707)资助。

摘　　要：零售企业的各项经营活动都离不开商品的销售情况,对商品的销售预测为企业制定生产计划与经营决策等活动提供重要的依据。针对企业销售额预测中销售额时间序列受外界条件影响大,预测精度低等问题,论文提出了一种基于GWO-Prophet的商品销售预测方法。基于某零售企业2015年-2018年销售额数据,通过Prophet模型将高维的销售额数据分别构建对应趋势项、季节项、节假日项、残差项的低维时序特征分量,分别用这些低维特征分量进行拟合后通过加法模型累加来预测未来一年的销售额数据;通过灰狼寻优算法(GWO)对Prophet模型参数进行智能寻优,防止模型陷入局部最优从而提高模型的精确度,通过灰狼寻优算法优化后的Prophet模型能更好地拟合突变点,季节项,节假日项等外界因素对销售额的影响。以MAE、MAPE和RMSE作为模型评估的指标,结果表明,基于GWO-Prophet模型的预测精度不仅优于单一的Prophet模型,还优于其他如ARIMA、SARIMA、LSTM对比模型。The business activities of retail enterprises are inseparable from the sales of goods.The sales forecast of goods pro-vides an important basis for enterprises to formulate production plans and business decisions.Aiming at the problem that the time se-ries of sales volume in enterprise sales forecast is greatly affected by external conditions and the prediction accuracy is low,this pa-per proposes a commodity sales forecasting method based on GWO-Prophet.Based on the sales data of a retail enterprise from 2015 to 2018,the Prophet model is used to construct the low-dimensional time series feature components of the corresponding trend item,seasonal item,holiday item and residual item through the high-dimensional sales data of the Prophet model.After fitting these low-dimensional feature components,the sales data of the next year are predicted by the addition model.Through the grey wolf optimization algorithm(GWO),the Prophet model parameters are intelligently optimized to prevent the model from falling into local optimum and improve the accuracy of the model.The Prophet model optimized by the grey wolf optimization algorithm can bet-ter fit the influence of external factors such as mutation points,seasonal items,holiday items on sales.MAE,MAPE and RMSE are used as indicators for model evaluation.The results show that the prediction accuracy based on GWO-Prophet model is not only bet-ter than the single Prophet model,but also better than other comparison models such as ARIMA,SARIMA and LSTM.

关 键 词：Prophet模型 GWO算法 时间序列 销售预测 可分解模型 

分 类 号：O141.4[理学—数学]