Summary: | 碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 92 === In software engineering it is very difficult but critical to forecast how much effort and time it will take to develop a system. The objective of this thesis is to provide a guideline to estimating the software development schedule when using the regression analysis techniques. The common causes of schedule variation will be identified first, and then be employed to construct and explain the software development schedule model.
We first use fishbone diagram and ANOVA approaches to identifying the common causes of schedule variation, and then establish the software development models in association with each individual factor. Next, we establish multiple-factor software development schedule model and give an example to compute the prediction interval. Lastly, we use the stepwise regression algorithm to refine the software development schedule model, i.e, removing the factor with prediction power but without contribution to the model. In addition, we also introduce nonlinear software development schedule model.
|