本课题以大型软件系统的软件维护为背景，展开"知识驱动"的软件维护关键技术研究，并据此开发可操作的多阶段融合的软件维护框架。该框架将软件维护经理的决策相关活动与软件维护人员的维护相关活动以多个阶段的形式有机地组织在一起，每个阶段的活动都由软件仓库中挖掘出的"知识"和前一阶段的信息驱动。该框架不仅能帮助软件维护经理选择合适的软件维护人员实施特定的软件维护任务，而且也能帮助软件维护人员快速理解所维护的系统并高效地完成维护任务。本课题的具体研究内容包括：(1) "知识驱动"的软件维护关键技术研究，包括维护作用域预测、维护工作量估算、维护任务智能分配、程序快速理解、变更自动推荐和易维护性度量；(2) 可操作的多阶段融合的软件维护框架开发；(3) 结合大型软件系统开展实验验证。本课题的研究不仅可以推动软件维护技术的理论研究，而且具有重要的应用价值。

This project first develops knowledge-driven software maintenance techniques and then uses them to build a software maintenance framework for large software systems. In this framework, the decision activities of a software maintenance manager and the maintenance activities of software maintainers are seamlessly organized in multiple stages. In particular, the activity at each stage is driven by the knowledge mined from software repositories and the information from previous stages. This framework can not only help a manager to allocate a given software maintenance task to an appropriate maintainer, but also help a maintainer to understand the system in a short time and to implement the task efficiently. The research contents of this project are listed as follows: (1) study of knowledge-driven software maintenance techniques, including maintenance scope prediction, maintenance effort estimation, maintenance task allocation, fast program understanding, automatic change recommendation, and maintenance measurement; (2) development of operational multiple-stage software maintenance framework; and (3) empirical validation of the proposed framework on large software systems. The outputs of this project can not only promote the development of software maintenance techniques, but also have an important application value.