A Pattern Recognition Approach to Software Engineering Empirical Modeling

计划、控制和预测软件过程的能力通过软件开发构件的度量来促进，尽管从这些构件构建定量模型是一项艰巨的任务。解释变量可能是众多和不同的；数据集充满了离群值；数据收集的质量可能令人怀疑；而且往往只有部分信息可用于建立模型。即便如此，仍需要定制的建模技术、程序和工具，以促进在特定开发环境中构建本地工具。开发过程中的三种活动需要随机建模：规划、监测和控制，以及事后分析和评估。模型用于使经理能够计划并采取纠正措施，以满足预算、进度和质量要求。这项研究的目的是继续开发一种称为优化集归约的方法，目的是解决软件工程的特定约束和需求。

The ability to plan, control and predict software processes is facilitated by the measurement of software development artifacts, even though construction of quantitative models from these artifacts is a difficult task. Explanatory variables are potentially numerous and heterogeneous; data sets are replete with outliers; the quality of the data collection may be questionable; and often only partial information is available for building the model. Even so, there is a need for customized modelling techniques, procedures, and tools to facilitate the construction of local tools within a specific development environment. Stochastic modelling is needed for three kinds of activity within the development process: planning, monitoring and control, and post-mortem analysis and evaluation. Models are used to enable managers to plan and to take corrective actions to meet budget, schedule and quality requirements. This research is aimed at continuing the development of an approach called Optimized Set Reduction with the aim of addressing the specific constraints and needs of software engineering.