存在偏離規格產出之兩階層最佳產品分配模式

Abstract

存貨管理與產品配置是營運策略及供應鏈管理中十分重要的議題，在無法確保生產品質的情況下，供應商為了減少成本，必須盡可能地將偏離規格的產品分配給消費者，同時滿足其需求。過往的研究通常使用機器學習將消費者對產品的接受度分成可以接受次級品及無法接受次級品兩類，然而這並沒有考慮到數量分配、產品性質及顧客購買紀錄。本研究參考了運輸問題中的路徑規劃模型，以此建立兩階層產品配置最佳化模型，試圖提供供應商最適合的產品配置模式以減少消費者退換貨的可能性。此外為了解決大規模整數規劃問題，我們引入了Benders演算法，利用切割平面的方式將模型分成主問題(以子問題的解為參數)及子問題(線性規劃問題)，並在過程中使用對偶理論進行求解。

Inventory management and product allocation play important role in operation strategy and supply chain management. Under the circumstances that most suppliers are unable to ensure the quality of production, they have to distribute the out-of-specification products to consumers as much as possible so as to minimizing the cost while satisfying consumers’ requirements. Previous studies usually use machine learning to classify consumers’ acceptance of product into unacceptable and acceptable substandard products, however most of them do not consider the distribution on number or quantity and the relationship between product characteristics and consumers’ purchase records. This study refers to path planning models in the transportation problem to establish a two-level product allocation optimization model, trying to provide suppliers with the most suitable product configuration mode to reduce possibility of returning and exchanging. Moreover, in order to solve the large-scale mixed integer programming, we introduce the Benders algorithm, which uses the cutting plane to divide the model into master problem (with the solution of subproblem as parameter) and linear subproblem, and make use of dual theorem in the process.