- 期刊
應用資料探勘分類法於多屬性ABC存貨分類
Applying Data Mining Methods on Multi-criteria ABC Inventory Classification
摘要
有效地管理存貨可降低存貨成本、提高供貨率，使企業競爭力提升。而針對物料加以分類，則是一項基本且必要的工作。分類不當則將造成後續採購與存貨管理的問題，進而造成企業的成本浪費。尤其是醫療服務單位的藥品及醫療耗材佔營運成本的比例極高。物料成本現今物料分類所依據的屬性，有別以往僅以物料之單位成本高低或年需求總金額作為分類依據，所考慮的物料屬性並非單一。目前有不少評估物料分類方法所採用的分析數據大都是根據Reid所提供的物料屬性數據，運用不同的多準則存貨分類模型針對醫院之47個存貨單位(SKUs)加以分類。雖然已有不少研究以多準則的分類方法應用在ABC存貨分類，然而，隨著資料探勘的技術逐漸受到重視，運用資料探勘的分類法可以解決多準則的計算問題，簡化計算以增進其實用性。因此，本研究乃採用Reid提供的數據，運用四種資料探勘技術(類神經網路、決策樹、貝式網路、隨機森林)進行物料分類實驗，並與使用此同一批數據的文獻結果相比較。研究結果顯示，本研究之分類準性不亞於其他文獻結果，尤其類神經網路方法之分類準確度達94.12%，因此資料探勘方法在在ABC庫存分類問題上兼具效率與實用性。因此本研究主要貢獻在於運用資料探勘的方法，提供快速有效高的料分類技術。
並列摘要
Effective inventory management can reduce stock cost and improve fill rate. The essential task is to conduct inventory classification. Especially, the cost of medicine and the materials dominate a large portion of the operating cost of the health care provider. Without a proper and accurate classification, it might cause serious waste. Due to the increasing complexity of the material purchased in the health care provider, the inventory classification is no longer based on only single criterion. Since the 1980s, there have been many studies on Multi-Criteria Inventory Classification problem which are based on Reid's published inventory data. Flores, Olson and Dorai presented a multi -criteria inventory classification model and applied it to classify 47 items of hospitals Disposable inventory units (SKUS). However, many of the above methods are too complex and difficult to use in practice. Therefore, this study is based on the data presented by Reid, using simple data mining classification method to achieve the classification results, and compare with the results and Flores, Olson and Dorai. The results show that the algorithms used in this study are no inferior to the results of the previous works. Especially, the percentage of correctness of Neural Network method is about 94.12%. The result of the proposed data mining technique can be applied to the case of multi-criteria inventory classification problem effectively and efficiently.