結合機器學習與統計預測之動態庫存管理

Dynamic Inventory Management through Machine Learning and Statistics Forecasting Methods

唐靚容(Ching Jung Tang)

指導教授 ： 吳政鴻

國立臺灣大學/工學院/工業工程學研究所/碩士(2017年)

https://doi.org/10.6342/NTU201703092

摘要

本研究旨在少量多樣生產模式下改善現有的庫存管理模型，由於少量多樣的產品容易受市場波動影響，具有不穩定的需求訂單，常常使庫存數量供需無法達成平衡，且產品種類豐富造成庫存管控較不易。故本研究將開發四種庫存管理模型，分別建立統計需求預測模型和安全庫存做為訂購的參考值，動態調整每一期的訂購數量，並期望以成本的角度選擇庫存風險較低的庫存管理模型，提供企業庫存管控的方針與決策的參考依據。
然而，目前企業對於少量多樣產品的作法是以「分類」的方式處理，根據不同的產品特性給定所屬的類別定義，並執行各類別對應的庫存管理策略，提升庫存管理的運作效率。透過機器學習方法對產品快速執行分類，學習訓練集合的特性並建立分類模型，接著再對測試集合進行類別定義，快速執行產品的分類，簡化少量多樣商品繁複的處理程序。

關鍵字

統計預測 ； 存貨管理 ； 動態規劃 ； 機器學習

並列摘要

The research aims to strengthen the recent inventory management model which focus on low-volume and high-mix production. The low-volume and high-mix production method usually couldn’t satisfy the demand on time due to the fluctuation of demand caused by market all the time. According to the high-mix product, it is hard to make material control decision clearly and quickly under abundant barriers.
Therefore, the research formulates four different inventory management model including statistic forecasting model and safety stock, and the cost is the point of view for each product to choose a better management model.
However, classification is a common method to manage the inventory in the practice of enterprise nowadays. It conducts different inventory management strategies by considering different categories of products. The categories are based on the product attributes and the operational efficiency. The research classifies rapidly learns the attributes of the training set and constructs the classification model by machine learning. The new model simplifies the complicated procedures of the low-volume and high-mix inventory management in practice.

並列關鍵字

Statistic forecasting ； Inventory management ； Dynamic programming ； Machine learning

參考文獻

[1] Alstrøm, P., & Madsen, P. (1996). Tracking signals in inventory control systems a simulation study. International Journal of Production Economics, 45(1-3), 293-302. doi:10.1016/0925-5273(95)00120-4

[2] Anderson, E. T., Fitzsimons, G. J., & Simester, D. (2006). Measuring and mitigating the costs of stockouts. Management Science, 52(11), 1751-1763.

[3] Bacchetti, A., & Saccani, N. (2012). Spare parts classification and demand forecasting for stock control: Investigating the gap between research and practice. Omega (United Kingdom), 40(6), 722-737. doi:10.1016/j.omega.2011.06.008

[4] Ballou, R. H. (2000). Evaluating inventory management performance using a turnover curve. International Journal of Physical Distribution and Logistics Management, 30(1), 72-85. doi:10.1108/09600030010307993

[5] Barankin, E. (1961). A delivery‐lag inventory model with an emergency provision (the single‐period case). Naval Research Logistics (NRL), 8(3), 285-311.