由於現今軟體環境之高度不確定性與市場的快速變遷，使得軟體專案時常被要求在更短的時間內完成開發。因此開發團隊常採取例如敏捷方式與彈性作法，省略系統分析設計等文件化工作而儘快進行實作，而這可能導致後續階段因為對系統的理解能力不足而有不良影響。軟體專案成員可以運用軟體反向工程的方式，在專案完成之後藉由程式碼分析產生相符的UML圖，得以解決前述問題。然而，專案若只藉著軟體反向工程工具自動化產生的UML圖，則可能有程式碼品質不佳連帶影響UML圖品質的情況發生，卻又缺乏一套驗證的機制。因此，本研究針對UML圖中的結構模型圖品質進行探討，以複雜度(Complexity)、耦合度(Coupling)、內聚力(Cohesion)和安全性(Security)作為品質評核指標，並發展各指標品質度量的公式。此外，本研究更進一步以類別圖為例並實作一個在軟體反向工程中針對UML類別圖的線上品質評核系統—Reverse Engineering Class Diagram Quality Inference System，旨在協助軟體專案成員藉由此系統進行軟體專案品質的評估，並透過系統所提供的品質評核評核指標與建議，進而改善專案以提升軟體專案品質。;Due to the high uncertainty in software environment and the rapid change in market, software projects are often required to complete in a short period. Therefore, the development teams often use flexible methodologies, omitting the system analysis and design work to start the system implementation as soon as possible. However, it may cause difficulties in later stages because the lack of understanding of system. To solve the problem and with object-oriented analysis design technique, we can use software reverse engineering to generate UML diagrams by code analysis after the system is implemented. If software projects only rely on the UML diagrams that generated by software reverse engineering tool, there may be poor quality of UML diagrams caused by poor code. However, such a problem cannot be known because the lack of verification mechanism. Therefore, the study focuses on structural UML diagrams and discusses the quality of structural UML diagrams, and uses the complexity, coupling, cohesion and security as the quality evaluation indexes. In addition, the study proposes an online system for evaluating the quality of UML diagram in reverse engineering and the system is focus on UML class diagrams, and is termed Reverse Engineering Class Diagram Quality Inference System. RECDQ aims at assisting software project members to evaluate the quality of the software project and improving the quality of the software project through the quality evaluation indexes and recommendations.
