A Study of Accuracy Improvement on Analogy-Based Software Cost Estimation Model
| Autor: | Nan-Hsing Chiu, 邱南星 |
|---|---|
| Rok vydání: | 2006 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 95 A reliable and accurate estimate of software development or maintenance cost has always been a challenge for both the software industrial and academic communities. Analogy-based method is a widely adopted problem solving technique that has been evaluated and confirmed in software effort or cost estimation domains. Similarity measures between pairs of cost drivers play a central role in analogy-based software cost estimation model. However, hardly any research work in the literatures has addressed the issue on how to determine the suitable weights of software cost drivers in the similarity measures of analogy-based software cost estimation model. And similarity, little theoretical or experimental works have been reported on the method of deriving a cost estimate from the adjustment of the reused cost based on the similarity distance. This dissertation aims to improve the estimation accuracy for the analogy-based software cost estimation model. A “weighted analogy-based software cost estimation adjustment model” is proposed in this dissertation. The proposed model utilizes the genetic algorithm (GA) method to determine the appropriate weights of cost drivers in the similarity measures of analogy-based software cost estimation model. In addition, the GA is also used to adjust the reused cost based on the similarity distances between pairs of projects. The approach of using two well-known software historical project datasets for verifying the proposed model is also illustrated in this dissertation. The experimental results show that applying GA to determine suitably weighted similarity measures of software cost drivers or applying a suitable linear model to adjust the reused cost is a feasible approach to improving the accuracy of analogy-based software cost estimation model. The proposed model is superior to regressions, neural nets or decision trees-based software cost estimation models. It also provides higher improvement rate on the accuracy of software cost estimates than those weighted or adjusted analogy-based software cost estimation model in the previous works. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|