|
During the evaluating of software quality, the common software quality model is abstract, and sometimes, comparation matrixes are inconsistent. In order to solve the problem, and therefore improve the pertinence, consistency and accurateness of software quality evaluating, this paper proposes a domain specific software quality evaluating method. The method contains constructing domain specific software quality evaluating model and optimizing fuzzy comparision matrixes.Firstly, we expound how to construct the domain specific software evaluating model. The universal model is the benchmark to software quality evaluating in all domains. On the one hand, select the applicable terms from GB/T16260. On the other hand, abstract the domain features from objective domain. Then, map the selected and abstracted terms to the universal model, hence produce the domain specific software quality evaluating model.Secondly, illustrate how to optimize comparision matrixes by Genetic Algorithm. The one iteration includes creating initial population, selection, crossover, and mutation. When meet the condition to stop, end the iteration. The optimized matrixes often have acceptable consistency. Then, work out the eigenvector of the maximal eigenvalue, and normalize it as a weight vector.Finally, based on the domain specific quality model and the optimized comparision matrixes, we evaluate several softwares in credit domain. We use.NET to realize the interfaces to users. Meanwhile, we achieve the Genetic Algorithm by MATLAB, ultimately we implement the quality evaluating module. The result in accord with the reality indicates that this method proposed in the paper is effective and practicable.
|