在以滿足顧客需求為競爭關鍵的時代中,如何建構一個有效的存貨控管機制將資源及存貨調配完善,及時將產品送到顧客手中完成交易,是企業提昇競爭力的重要關鍵。企業為了控管現今日益增加的客製化產品,若採用多物項存貨管理(Multi-Item Inventory Control)取代傳統的單物項存貨管理,可以有效節省成本與提昇服務水準。過去研究也針對存貨物項間的需求相關性進行探討,發現當物項間需求相關性越高時,採行多物項存貨管理節省成本的幅度也越大。然而以往學者所提出的需求相關性定義方法,常會隨存貨物項增加而增加定義的複雜性與困難度,使得多物項存貨管理的執行效益不彰。 有鑑於此,本論文提出一有效且客觀的關聯需求評量方法,以關聯規則(Association Rule)中的支持度計算作為衡量標準,取代需求共同發生的機率人為定義的缺點。然後對不同項目集的支持度進行距離函數轉換,再透過資料探勘的群集分析法(Clustering)對交易資料中的產品進行分群,使得每一項目群內之產品具有較為相似的銷售需求及特性,並將此群集結果應於多物項存貨管理模式中的可訂購策略(Can-order Policy),計算出各產品在不同時期所需的存貨數量後,採行共同訂購可以節省成本,同時也能降低因部分商品銷售過快,補貨不及造成損失的風險,達成創造利潤與提昇服務水準的目標。 為了驗證本研究所提出的方法為一有效的存貨控管機制,本研究透過數個範例資料,以模擬實驗的方式進行關聯分群方法應用在多物項存貨管理上的影響探討與參數分析。並以一實務上的案例資料,探討如何應用本論文之關聯分群方法,幫助該企業在實行多物項存貨管理的過程中,能持續有效地掌握顧客反應出的消費需求,適時提供適量的存貨補給,以達成節省成本與創造利潤的目標。
To satisfy customers’ requirements and increase competition in serve market, it is critical for an enterprise to construct an optimal inventory control model. For the sake of managing various costomized products, most enterprises adopt multi-item inventory control instead of traditional inventory management. A few researches have mentioned the coordinated replenishment policy with correlated demands reduces the corresponding total cost more than that with independent demands, and the savings increase as the demand correlation increases. However, this approach made the multi-item inventory control increase the difficulties for dealing with many products, so that the efficiency and benefits are usually not as satisfied as predicted. To conquer the difficulties, this research proposes an association clustering method to evaluate the correlated demands. The method adapts the concept of “support” from association rule algorithm to measure the similarity between products. After transforming supports for itemsets into distance function, we use clustering method to generate association clusters in which the items in the same cluster have similar demand models. These clustering results are used into can-order policy and calculating the quality of demand in different periods. Several simulation results show that the proposed method is efficient and cost-saving for several retailing transaction databases. In addition, this method also successfully applies to an enterprise’s transaction database to accomplish multi-item inventory control. It is found that the proposed method is beneficial for the enterprise to achieve a better and efficient inventory management.