A Study Of Intermittent Demand Combination Forecasting Based On Internal And External Data In The Context Of Intelligent Supply Chain

Intermittent demand forecasting is a long-term challenge in the field of intelligent supply chain,and accurate demand forecasting can improve inventory fill rate while reducing inventory cost and increase efficiency for enterprises to reduce investments.This study presents a series of intermittent demand combination prediction methods based on internal and external data.From the perspective of machine learning,this research constructs a complete intermittent demand feature engineering,predicting whether demand occurs through classification model,and predicting non-zero demand size through regression model.Based on the strategic selection of the inventory side and the stocking needs of the replenishment side,this study focuses on the optimization of the classification problem and proposes three optimization variants,which are stochastic perturbation optimization,optimal classification threshold searching,and transfer learning,respectively.Fusing the internal and external data of enterprises,oriented to the daily demand of spare parts inventory replenishment of vehicle enterprises,and based on the real auto after-sales business of traditional and new energy vehicle enterprises respectively.The data is tested,evaluated and validated in multiple dimensions.Compared with other intermittent forecasting methods,the three variants proposed in this study have been improved in terms of classification accuracy and forecasting precision significantly,which initially validates the potential of combined forecasting framework for intermittent demand forecasting and provides an empirical study of the framework in industry practice,which can further provide accurate upstream inputs for smart inventory and provide a dual guarantee and reference for enterprise smart supply chain decision making in terms of accuracy and efficiency.