資料倉儲之資料品質改善程序研究-以個案公司為例

Abstract

個案公司資料倉儲的建置方式是先建立各部門或各流程的資料市集（Data Mart）。當資料倉儲運作一段時間後，再新增其他主題的資料市集，以聚合而成企業的資料倉儲（Enterprise Data Warehouse, EDW）。而此狀況可能會影響到資料倉儲原有的設計架構，進而造成資料品質的不一致性。 為避免此狀況再發，同時也可以避免因人工協助驗證，而花費過多的人力成本在確保資料品質。透過資料評估的標準步驟，並改善步驟，以提升資料品質確認的速度，並延續驗證資料的經驗法則。同步建置資料品質監測系統，以取代人工確認資料品質問題，並減少確認資料品質問題的時間，也可快速反應資料問題點。 本研究採用四個構面為資料品質之衡量準則，精確性（Accuracy）、完整性（Completeness）、適時性（Timeliness）與一致性（Consistency）。探討透過資料品質改善流程，資料品質能有所提升。

We consider a company aiming to build a data warehouse. The company firstly establish some data marts of departments and processes. When the established data warehouse is operated for sometimes, and additional data marts of other topics are assembled into the enterprise data warehouse. However, this approach would affect the original design architecture of data warehouse and cause data quality inconsistencies. In order to avoid recurrence of this situation, additional effort of manual validation, and spend too much labor costs in ensuring data quality, this research, through the standard steps of data assessment, improves the way of promoting the speed of data quality verification and maintains the experience rule of data verification. The proposed method synchronizes the processes of building data quality monitoring system to replace manual data quality confirmation, reduces the time to confirm data quality, and expedite the data response. In this study, four dimensions of criteria measure are considered: data quality, accuracy, completeness, timeliness and consistency. Through the process improvement in data quality validation, data quality can be improved.