資料採礦在產品庫存價值分析之研究---以A公司為例

Application of Data Mining in Production Inventory Value Analysis A Case Study of A company

龔惠盈(Huei-Ying Gung)

指導教授 ： 張 百 棧

元智大學/工程學院/工業工程與管理學系/碩士(2005年)

摘要

近年來資訊科技成長快速及民國九十一年我國正式加入世界貿易組織(WTO)對於汽車零件產業將面臨激烈生存的競爭，而我國汽車零組件廠在顧客至上與賣方市場下也只能接受不同體系的供應鏈體系。
目前的企業生產情形，其產品型態大多已轉型為多種少量、生命週期與訂單交期日短、及採購與生產全球化的現象。供應鏈上游廠商為了因應終端產品的短生命週期及快速回應顧客，以致於造成各層廠商庫存的擠壓及需求放大，如此便使得製造過多的成品零件及物料的缺貨損失。因此在這競爭激烈的環境中，企業必須降低成本、提昇顧客滿意度、研發更新的產品來增加生存的競爭力。
本研究以顧客歷史交易資料進行產品庫存生產價值分析。首先採用Hughes 提出RFM 分析模型為基礎，增加以RFM 個別差異為權重值，將RFM 屬性轉換成三度空間向量，計算絕對距離做為產品價值指標。然後應用資料探勘(Data Mining)中具有學習能力的類神經網路(Artificial Neural Network)技術的倒傳遞類神經Back-PropagationNetwork，BPN〉 ，將過去期間的顧客交易資料做為輸入源，以目前期間的產品價值指標做為學習目標。經由學習所得的類神經網路模型，做為未來產品庫存生產價值指標的預測。最後結果可作為生管進行半成品庫存生產參考依據。

關鍵字

RFM 分析模型 ； 類神經網路 ； 資料探勘

並列摘要

ABSTRACT

Recently, the science and technology in Taiwan grow up fast and our country officially joined the WTO (World Trade Organization) in year of 2002, this made a great impact to the automobile parts industries. In order to survive in the intensive competition, the automobile spare parts factory in our country which is highly customer-oriented can only improve their efficiency of their supply chain system to adjust themselves to the seller’s market.
Making a comprehensive survey of the annual sales of the parts in our company, the types of products are classified into various items with the following characteristics such as low quantity, shortened product life-cycle, fast delivery time, and globalization in both purchasing and manufacturing. In supply chain system, the upper layer manufacturer in behalf of dealing with the short terminal product life-cycle and fast-responding customer service, so that to cause lower layer’s vendors magnifying the manufactory’s stocks extrusion and demand, thereupon making excess end items and materials to run out of stock.
In this thesis, we use customer history transaction data to perform production value analysis. First, based on the RFM analysis model and extended with RFM individual difference as weighted value, we transform RFM attributes into a three-dimensional vector to compute the absolute distance as the production value index. Then, with the Neural Network technique, which is a Data Mining technique with learning ability, we use past customer transaction data as input and current customer value indices as learning goal to produce a Neural Network model to predict future customer value indices. Finally, we actually perform computation and prediction of production value indices with customer history transaction data of two different industries.

Key word: RFM Analysis model 、Data Mining、Neural Network

並列關鍵字

RFM Analysis model ； Data Mining ； Neural Network

參考文獻

6.連惟謙，「應用資料分析技術進行顧客流失與顧客價值之研究」，中原大學資管研究所碩士論文，民國九十三年。

1. Spring and R. Boaden, “One More Time: How Do You Win

Orders?: a Critical Reappraisal of the Hill Manufacturing Strategy

Management, Vol. 17, No 8., pp. 757-779, (1997)

Priorities, Process Innovations and Time-Based Competition in the Manufacturing Sectors of Industrializing Economies: The Case of Turkey”, Benchmarking for Quality Management & Technology, Vol5, No. 4, pp. 304-316 (1998)