| As one of the most important tasks in judging the feasibility of software project, and in supporting to get project well planned, predicting software development schedule is a key task in software development. Moreover, alonging with the expanding of the scale and investment in software projects, software development schedule now plays a significant role. It's required to find out a proper, specialized method in accurately predicting the schedule. In software developing field, how to estimate the cost and schedule is a tough issue. This thesis initiates a regression method to tackle the problem by concerning with the factors influent schedule variation. By this method, we will extract those common factors and construct a schedule model to perdict and explaing the schedule. The major work and outcome are outlined as below:1, using fishbowl diagram and ANOVA method to extract the common factors which effects schedule, and construct the model by simple regression analysis method;2, leveraging multiple regression method to construct a software development schedule model which involves multiple factors, and specifies the method of predicting interval;3, advancing to a stepwise regression method to refine the model, to ease the duplication and overlapping of those predicting factors, and we also illustrate the non-linear schedule model.This thesis steps forward in software development schedule since we present a refined regression method, which is not considered by conventional models, and we also propose a non-linear regression model. Furthermore, we will continue to work in areas such as how to construct non-linear regression model, how to practice the method upon real projects, and how to estimate the viarations of different projects, or different phases in projects. |
