Integrate the GM(1,1) and Verhulst models to predict software stage effort
IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews
Data accumulation and software effort prediction
Proceedings of the 2010 ACM-IEEE International Symposium on Empirical Software Engineering and Measurement
| Hi-index | 0.00 |
This paper presents an empirical study of effort estimation. In particular, the study is focused on improvements in effort estimations, as more information becomes available. For example, after the requirements phase, the requirements specification is available and the question is whether the knowledge regarding the number of requirements helps in improving the effort estimation of the project. The objective is twofold. First, it is important to find suitable measures that can be used in the re-planning of the projects. Second, the objective is to study how the effort estimations evolve as a software project is performed.The analysis is based on data from 26 projects. The analysis consists of two main steps: model building based on data from part of the projects, and evaluation of the models for the other projects. No single measure was found to be a particular good measure for an effort prediction model, instead several measures from different phases are used. The prediction models were then evaluated, and it is concluded that it is difficult to improve effort estimations during project execution, at least if the initial estimate is fairly good. It is, however, believed that the prediction models are important to know that the initial estimate is of the right order, i.e. the estimates are needed to ensure that the initial estimate was fairly good. It is concluded that the re-estimation approach will help project managers to stay in control of their projects.