A general framework to improve software estimation

软件工程师准确评估项目的能力是软件管理学科中的一个重要问题。在软件工程中，已经有无数次尝试确定可以用来解决软件评估挑战的特定模型、工具或技术；不幸的是，大多数尝试都没有成功。另一方面，我们新的工作量和规模估计方法表明，神经模糊技术和算法模型的结合可以极大地提高软件项目估计的准确性。我们过去六年的工作结果证实，与单独使用算法模型相比，组合神经-模糊框架可以产生更准确的软件估计。我们的框架的一个最重要的特征是，它的体系结构本质上独立于算法模型的选择或预测问题的性质。

The ability of software engineers to assess projects accurately is a vital issue in the software management discipline. There have been countless attempts in software engineering to identify a specific model, tool, or technique that could be used to solve the challenges of software estimation; unfortunately, most of them have not succeeded. On the other hand, our novel approach to making effort and size estimations shows that a combination of neuro-fuzzy techniques and algorithmic models can greatly improve the accuracy of software project estimation. The results of our work over the past six years have confirmed that, as compared to the use of an algorithmic model alone, the combined neuro-fuzzy framework can produce more accurate software estimations. One paramount feature of our framework is that its architecture is inherently independent of the choice of the algorithmic models or the nature of the prediction problem.