考量價格與重要性因素下，需求分群與預測模式之研究-以汽車維修零組件為例

Abstract

近年台灣汽車市場持續增長，在2022年上半年台灣車輛工業達到新台幣3,607億元，顯示其在台灣製造業中的重要性。由於汽車屬於耐久財，消費者並非會頻繁更換汽車，因此對於品牌車廠而言，妥善的售後服務也成為關鍵的競爭因素，然而一輛汽車常由超過一萬個零組件組成，售後服務市場也非僅提供一種零組件，進而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，因此汽車零部件需求的理解和預測，以及庫存管理的妥善規劃，是提高零組件供應鏈的成功關鍵。 本研究的目的希望能有效藉由零組件的需求指標進行分群，藉此得到更多不同代表零組件特性與最佳預測方式間的關聯性，本研究使用汽車零組件的價格、重要性、平均需求間隔與需求變異係數作為分群指標，將具相似特性的零組件歸納為同一群，再套用於現行個案公司的需求預測方式以及本研究提出各預測方式於不同集群，找出各集群最適預測模型和集群特性之關聯。 研究結果顯示，在高單價與高重要性的零組件集群較適用於趨勢性的預測模型，而低單價高重要性的零組件適用於平穩的預測模型，以及低重要性的零組件較不適合進行時間序列法的預測，藉由找出零組件特性與適用預測模型的關聯性，在企業提供顧客售後服務時，能幫助企業根據特定特性的零組件，作出對應適合的預測方式與存貨規劃。

This study focus on effectively cluster the spare parts by the demand indicators of spare parts, thereby revealing more relationships between different representative spare parts characteristics and optimal forecasting methods. This study uses the price, criticality, and monthly demand volume of automotive spare parts as clustering indicators. Spare parts with similar characteristics are grouped together and applied to the current demand forecasting methods of the case company, as well as the various forecasting methods proposed in this study for different clusters. The goal is to find the most suitable prediction model for each cluster and its relationship with the characteristics of the cluster. The results show that the cluster of spare parts with high unit price and high criticality is more suitable for trended time series models, while the cluster of low unit price and high criticality spare parts is suitable for stationary time series models. Spare parts with low criticality are less suitable for time series prediction methods. By finding the relationship between spare parts characteristics and suitable prediction models, it can assist companies in providing after-sales service to customers. This study help companies to make appropriate predictions and inventory planning based on the specific characteristics of the spare parts.