A case study for software quality evaluation based on SCT model with BP neural network

With the increasing for function, scale, hierarchy and complexity of software project, the software life cycle and development stage show a trend of cross-cutting and fuzzy boundary. The non-technical factors, such as poor management and control during the implementation of software projects, are the major reason for causing the low success rate of software projects recently. Therefore, the software quality evaluation under complex environment should take the cross-influence between different stages of software life cycle and different quality evaluation standards into consideration. Our research is to construct a new software quality evaluation model by using the influence relationship and the influence intensity index between project management domain and project quality evaluation criteria including scope, cost, and time. First, we came up with the definition of software project management domain in the process of software project development and management. Second, we proposed a mathematical method for extracting the direct or indirect influence relation between them, and give a definition for the quantitative evaluation index and its calculation formula. At last we proposed to construct a neural network training model which includes evaluation model logic relationships and software quality quantitative evaluation index. Through study and training by simulated software project management data, we can discover some key data, such as normal threshold range of influence, factor weights, etc. Therefore, a complete evaluation system is built, and the scientific nature and accuracy of the proposal evaluation system will be improved.