花卉資料倉儲的三種改進方案：權限控管、審計資訊與季節性資料分析

Abstract

資料倉儲在管理上有三項主要的挑戰，分別是身份驗證與授權、稽核追查及應用。本論文是針對台灣花卉資料倉儲的每項挑戰提出改進方案。就身份驗證與授權的問題而言，是以雙主機的硬體架構來分擔資料倉儲龐大的資料負荷。而系統資料負荷的減輕能夠縮短線上分析處理所需的時間。在稽核追查問題的部份，由於無法從資料倉儲直接取得審計資訊。因此將資料倉儲的審計日誌轉為審計資訊，並以審計資料庫為基礎建構一個三層式架構的網站。使用者可以藉由網頁追蹤查詢自已的使用記錄。最後就應用的部份，是以時間序列的方法並考慮季節性來分析花卉交易資料。而季節因素的考量可以增加對花卉資料預測的準確性。透過上述三項改進方案能夠提升花卉資料倉儲的效能並使系統的運作更有效率。

There are three major challenges in the management of data warehouse: authentication/authorization, accounting/auditing, and application. This thesis proposes an improved method to each challenge for the data warehouse of flower industry in Taiwan. For the authentication/authorization problem, a two-layer hardware architecture has been implemented as the platform of data warehouse for sharing its huge data loading. The reduction of data loading makes the processing time of On-line Analytical processing (OLAP) shorter. For the accounting/auditing problem, it is impossible to compute the payments of users from the data warehouse directly. So a three-tier web architecture has been set up for transforming the audit log in the data warehouse into a payment schema in a data base. Also users can trace their accounts through web pages. For the application problem, a seasonal time series method is proposed for analyzing the flower transactional data. The consideration of seasonal factor makes the forecasting of flower transactions more accurate. The three improved methods make the operations of flower data warehouse more effective and efficient.