透過您的圖書館登入
IP:18.118.99.234
  • 期刊
  • OpenAccess

以映射技術拓展企業資料庫之研究

Extending Enterprise Database by Mapping Technology

摘要


在資料庫分析的過程中,若資料庫不完整或不正確,將無法提供正確且有貢獻的知識。本研究將以映射技術嘗試從輔助資料庫映射目標資料庫所欠缺的欄位到目標資料庫中,使目標資料庫之欄位更完整,以利後續分析利用,提昇資料庫之價值。本研究架構包含三大部分,第一部份為前置處理,此部分是系統裡唯一需要人為處理的部分,此部份需要將資料之缺失值與錯誤值予以插補及修正,並且將兩資料庫之欄位格式統一;第二部分為關連建立模組,此部分將針對主觀認定的虛擬索引欄位進行統計檢定,以判別是否有統計上的可信度;第三部分為映射模組,希望藉由統計檢定的虛擬索引為關連,以線性網路、機率類神經網路、徑向基網路、倒傳遞類神經網路四個分類方法,進行分類模型的訓練,最後挑選此四個模型中準確度最高者為映射模型,將欲映射欄位映射到目標資料庫中。本研究之系統實驗採用工商及服務業普查資料庫為目標資料庫,以技術創新調查資料庫為輔助資料庫,再運用本研究所提之技術將兩者共有欄位由後者映射到前者中,並衡量映射值之準確度,以證明依本研究所提出之系統架構建立的拓增欄位有其仿真性。

關鍵字

資料庫 映射 分類 類神經網路

並列摘要


We all know that information is an invaluable competitive capital. This research addresses a mapping framework to expand a target database by mapping from an auxiliary database. The mapping framework includes 3 parts. The first part is the preprocessing module. The 2nd one is the relationship construction module. It uses M-W U test to determine virtual indices. The 3rd one is the mapping module. It chooses the best result among the results from LN, RBF, PNN, and BNN mapping from the assistant database to the target database. The experiments use real data and use our mapping framework to map common fields from the latter to the former. Finally, it proves that expanding fields which are mapped by the mapping framework have high similarity to real values.

並列關鍵字

Database Mapping Classification Neural Network

參考文獻


Estefane, D. L.,Andre, D.C.(1999).Credit Analysis Using Radial Basis Function Networks.3rd International Conference on Computational Intelligence and Multimedia Applications.(3rd International Conference on Computational Intelligence and Multimedia Applications).:
Fayyad, U. M.,Piatesky-Shapiro, G.,Smith, P.,Uthurusany, R.(1996).Advances in Knowledge Discovery and Data Mining.Cambridge:The AAAI Press.
Frawley, W. J.,Paitetsky-Shapiro, G.,Matheus, C.J.(1991).Knowledge Discovery in Databases.California:AAAI.
Golub, G.,Kahan, W.(1965).Calculating the Singular Values and Pseudo-inverse of a Matrix.SIAM Journal of Numerical Analysis, Series B..2,205-224.
Han, J.,Kamber, M.(2001).Data Mining: Concepts and Techniques.San Francisco:Morgan Kaufmann.

延伸閱讀