台灣高速鐵路營運資料倉儲建構之研究

Abstract

台灣高鐵於2006年11月1日起正式營運後，各類如行車控制，訂位票務等交易型營運系統，由於其系統架構、導入方式、建置廠商、發展時程不盡相同，不但造成未來系統間之資料交換及連結作業疊床架屋、維護困難，也無法在未來營運累積了大量歷史資料後，對公司提供整體性之經營績效評估及決策資訊。因此一個完善的系統整合機制，勢必要提前規劃，屆時才能發揮資訊的最大價值。而「資料倉儲」(data warehouse)及「線上分析處理」(on-line analysis process)理論，可將企業大量資料作最有效管理，以呈現動態的決策性資訊，正提供了一個完整的解決方案。 本研究假設，企業需求引領了資料倉儲的data model，而以支援企業策略為最終目標。因此本研究以「分析、設計、實作」三個階段，逐步建構以平衡計分卡之績效衡量為基礎，「關鍵績效指標」(KPIs)為需求之資料倉儲系統：(1).在分析階段，「企業需求」(Output)部份由BSC四個構面的策略目標歸納出八個績效分析主題；「資料來源」(Input)部份篩選出與營運績效計算最相關的三個子系統。(2).在設計階段，推導出八個維度(dimension)及六個事實(fact)資料表的多維度資料倉儲模型，並完成其與來源資料的對應轉換關係。(3).在實作階段，將模擬了兩年度的營運歷史記錄，以工具實際萃取、轉換、載入至資料倉儲後，建立了8個資料方體(cube)，並以網頁方式呈現了績效指標之線上分析操作。 透過整個研究過程，顯示結合了平衡計分卡績效衡量架構，以及資料倉儲的分析工具及技術，確能達到績效管理的自動化，快速反應營運關鍵性指標，提供即時、正確的決策支援，並也提供了企業一個更容易成功建置資料倉儲的模式。

When Taiwan High Speed Rail (THSR) formally starts its operation on 1st Nov 2006, various kinds of transactional operation systems such as Traffic Control, Reserving and Ticketing, etc., due to their different system structure, implementation method, structuring company, developing schedules, not only the data exchange and linkage among the systems is complicated, the maintaining is difficult, after the accumulation of large amount of archive data, it is unable to provide data for overall operational performance evaluation and strategy making. Therefore, to enable data for it mass value, it is necessary to plan in advance a flawless system integration mechanism. Data Warehouse and On-line Analysis Process theories can most effectively manage the mass amount of data of corporation, to display an active strategy-making data and to provide an overall solution. This research, of which the final target is to support the corporation’s strategies, hypothesizes that corporation’s request lead the data model of data warehouse. Therefore this research constructs a data warehouse system in three phases - analyze, design and practice, with the basis on performance evaluation of Balanced Scorecard, and the need of Key Performance Index(KPIs). (1).In analysis stage, “Corporation’s need”(Output) is summarized to 8 performance evaluation subjects according to 4 BSC strategic targets. “Data Resource”(Input) is selected to three sub-systems, which are mostly related to operation performance calculation. (2).In design stage, a multi-dimensional data warehouse model is developed with 8 dimension and 6 fact tables. Also the exchange with the resource data is completed. (3).In practice stage, this research simulates two’s archive data, actually extract, transfer, load data onto the data warehouse and build 8 data cubes, and at last, displays on-line analysis of performance index in web page. Through out the whole research process, performance evaluation structure of BSC and tools and technique of analysis of data warehouse are combine to achieve a automatic performance management, which can rapidly reflect key operation index and provide fast and accurate strategy making support. It can also provide the corporation an easier way to successfully build up a model for data warehouse.