現代軟體不斷的演進，雖然在版本控制系統中有詳細的軟體版本異動資訊， 但是對於日益複雜的軟體結構的理解仍是有限的，同時，一個專案動輒有數千甚 至上萬行程式碼，所以使得軟體維護成為一項難題。目前對於版本控制系統的資 料挖掘研究多使用關聯規則，常會忽略可能存在的程式異動順序關係，因此本研 究加入時間維度，從網路上開放的CVS程式庫取得資料並利用循序探勘技術加以 分析，找出潛在的序列樣式，與先前相似的研究相比能更精確的指出軟體專案需 變動的「個體」，並以一個分類規則檢視序列樣式對於使用者的價值，以提供未 來軟體維護的參考。 The evolution of the modern software is continual. Although detailed information of the evolution of the software version is stored in the version control system (VCS), the understanding of more and more complex software structure is still finite. On the other hand, lines of code in the software project are usually hundreds of thousands, which makes the software maintenance a difficult problem. The researches of applying data mining techniques to VCS are usually based on association rules, which usually pass over the ordering information. As a result, by taking the time dimension of the software data from the Concurrent Version System (CVS) into consideration, this study uses the sequential-pattern mining technique to analyze and find out the potential sequence pattern. We expect the “entities” to be changed more precisely than those in previous researches and re-evaluate the sequence pattern for the users by means of a classification rule in order to provide the reference of the software maintenance in the future.
