以全球資訊網為基礎之ERP資料探勘整合系統

Abstract

在企業相繼完成交易流程(Transaction Processing)的自動化之後，決策流程(Decision Processing)以及預測流程(Forecast Processing)的自動化將是企業未來競爭力的關鍵所在，為了彌補ERP系統在預測流程方面的不足，所以在此提出一套架構於ERP系統下之可自行建構符合本身企業需求的、即時正確的以及可預測未來市場變化的資料探勘系統架構。 本研究所提出的架構是以ERP強大的交易流程處理能力為基礎，利用資料倉儲技術以及OLAP技術將ERP的交易流程和決策流程整合，再透過資料探勘技術將決策流程與預測流程整合，資料探勘引擎則是實際設計出一個分類的模型，因為預測流程和ERP之自動化交易流程緊密結合，所以本研究結果能夠將傳統資訊系統需要3周的模擬分析，在1個小時內分析並且預測完成，讓企業主在做決策時游刃有餘，足以因應瞬息萬變的市場，立企業於不敗之地。 本研究將ERP資料探勘系統應用於IC測試業，利用本研究之ERP資料探勘系統來探勘測試機台的測試狀況，針對造成測試機台當機次數過高的因素予以解決，根據實際驗證得知其測試機台當機次數平均改善率為83%、測試異常次數平均改善率為93%以及人為失誤次數平均改善率為96%。

After most of the enterprises have finished one after the other in the automation of transaction processing, the key issue for the future competitiveness of enterprises shall lie on the automation of decision making processing and forecast processing. To complement the deficiency of the forecast flow in ERP system, we propose here a data searching architecture that can be built on the ERP system while meeting the individual needs of enterprises, providing real time accurate forecasts on changes in the future market. The architecture proposed in this study is based on the potent transaction flow processing power of ERP. It integrates the database and OLAP technologies, transaction and decision making flows of ERP, then uses data searching technology to integrate decision making and forecast flows. The data-searching engine is the actual design of a categorizing module. As the forecast flow is closely integrated with the automated transaction flow in ERP, this research can finish within an hour the analyses and forecasts, which the traditional information technology system would take about 3 weeks to process the mock analyses. This makes entrepreneurs much easier in decision making and thus enables them to well handle the always-changing market and establish themselves firmly. This research applies the ERP data searching system on the IC testing industry. Using the ERP data searching system in this study to investigate the testing status of the testing machine, resolutions are found pinpointing the factors that causes the too high frequency of machine down. According to the actual experiment, the tested testing machine shows an average improvement rate of 83% over the number of machine down, average improvement rate of 93% over testing abnormalities, and average improvement rate of 96% over the men’s error.