花卉產業的第二次台北資料超市再設計以解決其維度緩慢改變問題

Abstract

維度緩慢改變問題是指在多維度資料超市或倉儲中，部份維度資料的意義有緩慢改變現象。本論文針對花卉產業中台北花市的維度緩慢改變問題進行研究。由於花卉資料的意義源自變動快速的花卉產業環境，研究結果顯示維度緩慢改變問題將必然存在，且應持續投入維護成本以修正此問題；而不同維度的改變速度也不盡相同，改變速度越慢的維度將越難察覺此問題。為解決現階段發現的特定緩慢改變維度問題，本論文利用錯誤的統計資料為基礎，對資料提出系統性判斷與修正方法。為了證明此方法的可行性，利用SQL Server整合服務實作ETL系統。

Slowly Changing Dimension (SCD) problem in a multi-dimensional data mart or warehouse refers to the phenomenon that some dimensions owns data with a slowly changing meaning. In this thesis, the SCD problem of floral industry in Taipei has been studied. Since the meaning of floral data evolves with the dynamic development of the floral industry, the research results show the inevitable of the SCD problem and the necessity of a persistent maintenance cost for fixing its dimensions. Moreover, each dimension has its own speed to change the inherent meaning, and a more SCD is harder to be found than a faster one. Accordingly, for solving the specific SCD problem we identified at the current stage, a systematic diagnosis method based on the faulty statistic we have known is proposed to locate the troublesome dimensions. Also for showing the feasibility of this diagnosis method, an Extract-Transform-Load (ETL) software based on SQL Server Integration Services (SSIS) has been coded.