Research And Implementation Of Software Cost Estimation Method Based On Constructive Cost ModelⅡ

As software enterprises are increasingly in demand on process capability, software cost estimation which is one of the important methods of improving process management is gradually concerned by them. Effective software cost estimation is the premise of reasonable control and scientific management of software products. At present, software estimation methods are relative mature in foreign countries. But software estimation is a weak link in domestic project management. In those few published models, COCOMOII (Constructive Cost Model) is a relative widely used and mature model. So proceedding from domestic software environment, how to identify factors which have more effect on the model and how to impove the model to adapt to the domestic environment is very valuable. The main analysis and research works of this paper are as follows:1. Combined with domestic software industry, analysised concept and value of every cost factor in COCOMOII, established the objectives and principles of improvement, improved cost factors localizedly, and provided a new formula for effort estimation. Localized improvement reduced the complexity and difficulty of application of the model. So it saved itself workload of cost estimation.2. Collected historical project data in some enterprise, adopted linear regression method and used the data processing software of least square method to calibrate parameters in the new effort estimation formula. The formula whose parameters have been calibrated was more appropriate to the enterprise's environment and helped to improve the accuracy of estimation.3. Because function point could convert to lines of code as size input of the localization effort estimation formula, this paper improved function point method of size estimation. Used use case model modeling technology and class diagram of UML to standardize the process of function point estimation and provided mapping rules. This improvement provided guidance for the process of function point estimation. At the same time, it to some extent alleviated the problem of decomposing function randomly which is caused by non-standardized document in local enterprises.4. Applicated model to four modules of Office Automation System in some enterprise to estimate effort. These modules are in similar development environment. Then compared the effort which was estimated by using new model with actual effort and compared the effort which was estimated by using original COCOMOII with actual effort respectively. The evaluation results showed that the model application effect is good.