Study On Evaluation Model Of Software Requirement Analysis Risk Based On BP Neural Network

With the rapid development of computer technology, software underwents a dramatic transformation. But the phenomenon of project delay and out of budget is widespread. Due to the disordered communication and changed requirement in the beginning of software development, the software requirement departs actual demand and causes the risk of software requirement analysis. Software requirement analysis is an very important phase in the software development process. The cost of repair defects in software maintenance phase is 100 times if we don't find and cnotrol the risk effectively in software requirement analysis phase. So, if we want to reduce development costs, meet user requirements, we must control risks effectively in software requirement analysis phase.At present, scholars have put forward many management models of software risks, but few study the software requirement analysis risk. Evaluation index of few models are not comprehensive when evaluating the risk of software requirement analysis; there is no combination with quantitative knowledge and qualitative knowledge in the methods of evaluation. Therefore, it is of great significance for the study that how to ascertain the potential risks and control them effectively,we put forward an evaluation model of software requirement analysis risk based on neural network to solve the problems above and explored from the following aspects:First, according to the factors in software quality, we create a component attribute mapping, which evaluates the creditability of components description from 17 aspects, designed to enhance the applicability of the creditability.Secondly, at the basis of the risk assessment indexs, we propose an evaluation model of software requirement analysis risk for forecasting the risks of requirement analysis, which combine Fuzzy Logic with BP neural network. According to the empirical knowledge of experts, we set the risk levels, which are the inputs of BP neural network after the fuzzy comprehensive evaluating. Finally, we obtain the risk levels. Third, we create a database of history risks, which will track and record risk information of software requirement analysis, and provide history data for evaluator.Finally, through simulation experiments we verify that the evaluation model of software requirement analysis risk proposed in this paper is feasibility and effectiveness.