大部份的軟體在開發完成後，會因為使用者需求的改變、軟硬體設備更換、軟體缺失的修復及增加系統彈性等原因，或者為了應付未來環境的變動，而持續的進行軟體的修改工作，因此軟體的維護是軟體生命週期中不可忽視的一環。為了確保軟體維護專案成功與修改後軟體的品質，軟體維護工作量預估的精確性，是一個很重要的因素。以往針對軟體維護工作量預估的文獻，很少有著重與區別不同維護的類型，因為各種不同類型軟體維護性質的差異很大，影響維護工作量的因素也有不同，因此軟體維護工作量之預估若能將這些不同類型之維護專案分類來建置預估模式，模式預估值的精確性應該可以提昇。
本研究即是在上述的研究動機下，探討依軟體維護類型將專案歷史資料來分類以建置個別之工作量預估模式，再將個別之不同類型之工作量預估模式所產生的預估值加總，與未分類所建置之工作量預估模式所導出之維護工作量預估值做精確性的比較，並探討不同維護類型所需之工作量影響因素是否有不同。本研究採用ISBSG軟體維護與支援資料集，將這些資料集內的歷史專案資料依四種軟體維護類型：完備性、預防性、更正性、調適性來分類，探討資料集之軟體維護專案的各項屬性，例如專案類型、應用系統類型、開發方法、生命週期模式等，對各種類型的軟體維護所需之工作量是否會有顯著影響，在採用統計迴歸方法分別建構各種軟體維護類型所需之工作量預估模式。另本研究也將軟體維護歷史專案資料依照不分類的方法來建置工作量預估模式。經由本研究的結果發現，各種不同軟體維護類型的工作量影響因素子確實有差異，然而在分類與不分類工作量預估模式之精確性卻沒有顯著的差異，因此本研究除了推論可能的原因外，並對後續的研究提出建議。

After the development process, most software will be modified based on user requirements changes, software/hardware changes, software defect repair and system functionality enhance or some other reasons such as environment changes during the software maintenance life cycle. In order to ensure software maintenance success and the modified software quality, the accuracy of software maintenance effort estimation is a key factor. The past literature on software maintenance effort estimation rarely explored the impacts of classifying software maintenance projects on the accuracy of the established software maintenance effort estimation models, because the differences of each type of software maintenance types are significantly apparent and their affecting factors of the required software maintenance effort are also different. Therefore, it is intuitive that the accuracy of the established effort estimation model could be improved if the models are established from the classified type of software maintenance projects.
Based on the above motivation, this study investigates the impact of classifying the software maintenance projects on the accuracy of the established maintenance effort estimation models and further compares the difference of effort factors of the models established by unclassified and classified software maintenance projects. This study utilizes the ISBSG software maintenance & support dataset, which the maintenance projects are classified into four types: perfective, preventive, corrective and adaptive, and investigates if the impact of the software maintenance effort estimation models established by using a statistic regression method and classified by different attributes of the maintenance projects, such as project type, application type, development methods, and life cycle models, on the accuracy of the established models are significantly different. The empirical experiment results show that the affecting factors of each of software maintenance project types are significantly different. However, the accuracies of the estimation models established by classified maintenance types and an unclassified type are not significantly different. Accordingly this study not only inferences the possible reasons but also makes recommendations for the future research.

[1]林信惠，「軟體專案管理」，智勝文化，民國91年1月
[2]吳宗正，「迴歸分析」，三民書局，民國82年5月
[3]顏月珠，「商用統計學」，三民書局，民國69年10月
[4]吳明隆，「SPSS與統用分析」，五南圖書，民國94年5月
[5]林傑毅，「整合風險評估之軟體專案預估模式」，碩士論文　國立台灣科技大學資訊管理研究所，民國91年
[6]許兆榕，「多因子資料分群對軟體工作量預估值精確性影響之研究」，碩士論文　國立台灣科技大學資訊管理研究所，民國93年
[7]劉亦真，「軟體專案資料分類對工作量預估值精確性之影響」，碩士論文　國立台灣科技大學資訊管理研究所，民國92年
[8]Piazza Roma,“Early Effort Estimation of Massive Maintenance Processes”
[9]Yunsik Ahn,“The Software maintenance project effort estimation model based on function points”,Journal of Software Maintenance and Evolution(2002/07)
[10]IEEE Standard “IEEE Std 14764-2006 Software Engineering—Software Life Cycle Processes—Maintenance” the Institute of Electrical and Electronics Engineers (2006/09)
[11]International Standard “ISO/IEC 14764 Software Engineering—Software Life Cycle Processes—Maintenance” International Organization for Standardization (2006/09)
[12]UKSMA, “Measuring Software Maintenance and Support, Version 0.5, Draft, July 1st, 2001,”United Kingdom Software Metrics Association
[13]Bennet P. Lientz, “Problems in Application Software Maintenance”Communications of the ACM(1981/11)
[14]ISBSG, “Data Collection Questionnaire Application Software Maintenance and Support,”International Software Benchmarking Standards Group,(2007/09)
[15]Bilge Baskeles,“Software Effort Estimation Using Machine Learning Methods”IEEE(2007)
[16]Frank Niessink,“Two Case Studies in Measuring Software Maintenance Effort”,Faculty of Mathematics and Computer Science()
[17]Kristian Marius,“Increasing Software Effort Estimation Accuracy Using Experience Data, Estimation Models and Checklists”Seventh International Conference on Quality Software(2007)
[18]Lientz, B.P., and Swanson, E.B. “Discovering issues in software maintenance. Data Management”, 16, 10, (Oct 1978), 15-18.
[19]Magne Jorgensen,“Experience With Accuracy of Software Maintenance Task Effort Prediction Model”IEEE Transactions on Software Engineering,(1995/08)
[20]Victor Basili,“Understanding and Predicting the Process of Software Maintenance Release”
[21]W.A. Delaney, “Predicting the costs of computer programs,” Data Processsmg Magazine, pp. 32-34,1966.