考量銷售量下之ARIMAX需求預測模型–以汽車維修零組件為例

台灣的汽車市場至今已發展成熟，對車廠而言，妥善的售後服務也成為關鍵的競爭因素之一，然而，汽車乃由一萬多種的零組件組成，不但因此形成了複雜的零組件供應網，對於車廠在存貨管理上也帶來了許多難度，其必須經由與國內協力廠商、國外原廠訂貨，在維持一定的服務水準下，透過存貨模型盡可能控制缺貨、庫存成本，以及供給、需求的不確定性，而存貨模型中的需求預測，在降低需求不確定上即扮演著關鍵的角色，良好的預測模型，將能有效地幫助控管零組件的進貨量，使整體營運效率有所提升。本研究主要目標在於汽車零組件的需求預測，在汽車保固年限內，藉由顧客可能的回廠維修、保養行為作為基礎假設，將銷售端的車輛銷售資料與售後零組件的需求資料進行整合，建構加入外生變數的ARIMAX時間序列模型，並與單純依據售後零組件需求建立的ARIMA模型進行預測成效之比較，藉此展現跨部門的資訊流整合，將有助於增進公司在存貨管理上的效率。由需求預測比較結果顯示，本研究使用的外生變數，即汽車車輛銷售量，能在特定的零組件上有效的增進ARIMA預測的表現，而在使回廠行為的假設下，部分零組件因其需求模式、零件特性的不同，在預測成效上也有所差異。本研究提出之方法，提供車廠於售後零組件需求預測上新的概念，透過車輛銷售、顧客回廠紀錄、歷史零組件需求資料，讓企業能針對零組件存貨系統的管理上，透過公司內的資訊整合，進行可能的優化與改善。

關鍵字

汽車銷售量；汽車零組件需求預測； ARIMAX

並列摘要

This study focuses on developing a forecasting model for the demand of auto spare parts. Under the warranty period of the car, we assume that the customers are willing to return to dealer for repairing and maintenance service. Based on this assumption, this study integrates the auto sales data into a forecasting model of predicting the demand of spare parts. Our model constructs an ARIMAX time series model with exogenous variables, and compare the forecasting results with an ARIMA model based solely on spare parts demand. The comparison results of demand forecasts show that the exogenous variables used in this study, that is, the sales volume of automobiles, can effectively improve the performance of ARIMA forecasts on specific components. Under the assumption of returning to the dealer, some components having different demand patterns and part characteristics lead to different forecasting results. The method proposed in this research provides new ideas for spare parts demand forecasting by using auto sales quantity data. Finally, our study shows that through combining vehicle sales numbers, customer returning records, and historical demand data of spare parts, companies can manage the component inventory system through the company’s internal information integration for possible optimization and improvement.