Research On Estimation Model Of Software Cost For Electric Power Industry

The domestic power industry software project estimation is still stuck in theexperience phase, and seldom use model approach. COCOMOII is a matureinternational and influential model. However there are22input variables, the model iscomplicated, and not easy for practical usage.Domestic and international environmentof software development and management level differ, it will lead to considerableerror if the model is used directly in the domestic project estimation.Through analysis, this article presents the idea of the improved model for thepower industry, but the difficulty lies in how to customize the model for the powerindustry. The majority of factors of the COCOMOII model are subjective factors, andwe have to try our best to lower such subjectivity in our improvement of this model.Hence, during research, learning from the expert approach which is used for thevariable preparation by the model’s original author, we invite26power industryprofessionals to assign values to the22variables of eight projectsof his specialty.Then we analyze the data in thesurvey questionnairesthrough weighted statisticmethod to choose the values and factors. Through this method, we lower thesubjectivity to a center extent, and decrease the number of factors from22to11.Meanwhile, the model constants are adjusted by applying the model to5cases, andthen customize them for the industry.The customized model is assessed by comparing with the actual value of3cases,the average error of workload estimation is improved and goes down to17.24%compared with the original model’s92.12%. The average error of the timeline isimproved and goes down to15.50%compared with the original model’s91.55%. Theimproved model meets the industry’s goal in terms of accuracy and simplicity.A customized software tool for the power industry is also developed based on theimproved model, and is being used in the project, based on user’s feedback, itimproved a lot compared with experience estimation approach, also the tool is simpleto use. After the improved is used in the actual projects, experience can beaccumulated, as a result, the model will be improved further.