發展基於使用者行為導向之智慧型財經資訊系統

Abstract

有鑑於全球投資環境快速變遷，投資市場趨向於全球化，各種投資資訊不分畫夜從各地湧流而來，而網際網路的推波助瀾更加速了資訊的傳遞與投資的方便程度。國內各大券商紛紛提供了線上下單服務，以財經資訊為主的入口網站相繼設立，投資人可以很快速便利的獲取財經資訊，經由電子下單系統迅速完成投資理財；這樣的環境讓每個人都有機會成為投資人，運籌於惟幄為自己創造財富。 在便利了投資管道與豐富財經資訊後接踵而來的問題便是資訊過量，投資人感受到壓力，過多的資訊提高了決策的複雜度和決策時間的延遲，對投資人而言反而是負擔。 本研究以解決資訊過量問題為研究目標，擬發展出具有資訊客製化能力的財經資訊系統解決方案。透過使用者瀏覽記錄的分析，使用適合的資料採礦模式萃取使用者行為偏好用以建構行為資料庫。雛型系統以代理人為中心，使用Gerard Salton所發展出來的Vector Model資訊檢索技術處理財經資訊的分類，截取行為資料庫內的使用者偏好做為重組網頁資訊的參考依據。 研究過程中考量了財經資訊的特質而設計，發展出來的流程可以用來做為處理財經資訊的模式，除了能夠達成資訊客製化的要求外，現有電子券商下單系統不需大幅度的修改，只需要加入代理人系統便可達成客製化的服務，對電子券商來說算是廉價又具效益的解決方案。為求客觀，本研究引用現有使用中的財經資訊系統，分析已有的使用者之瀏覽行為，並且在現存的系統上再做延伸來達成客製化資訊系統的要求。經過實際的建構測試後，發現此一客製化資訊處理流程的確適合作為現有財經資訊系統實行客製化服務的參考。

For the past few years, the financial investment run to diversification. An investor has many choices to invest. Meanwhile market is trade to worldwide, Investors have no choice to face complex and fast investment environment. In addition, today, the WWW is the most popular system in the Internet, more and more information distribution and acquisition rapidly, the broker must support all kinds of information for users to make decision trade financial product efficiently. Because financial investment diversification, market worldwide, and Internet system’s popular cause information explored, people can’t make decision easing due to too much information to be digest. Besides investors like to that personal style of information presentation in their monitor, current Internet based financial information system is not satisfied for them. In order to solve above problems, in this thesis an intelligent financial system based on user behavior is proposed which is allowed that the system not only can catch personal browsing behavior but also filter and sift news for an individual. Data mining methodology, include browsing sequence and association rule, and automatic text indexing to construct behavior database, and to use IR to filter news are also provided in this system.