A Software Size Estimation Method Based on Improved FPA

Software size eatimation is the key of entire software program project, and the accurate estimation immediately affect the success of project. An improved function point analysis (FPA) method was proposed for analyzing the software size. The method combined the advantage of the fuzzy rules and back propagation (BP) network. Firstly, fuzzy inference system based on the complexity weight matrix of function component was estabilished. Then the adjusted complexity weight were used for modifying the software function point. The adjusted data as samples were tranfered to BP network. By the advantage of BP network function approaching, the relationship between software components and software size was establish. Finally, BP network was used for estimating software size. The experiment results show that the method could eliminate discontinuity among the different complexity grades, and could make the best of history data, which enhances the accuracy of function point estimation.