With the improving of software technology and the level of project management, the success of software project is not a special case. But this does not mean that software project can be finished easily. In fact, the success ratio of software project is still low. The existing of Software system aim to reduce the uncertainty of practical system, but the process of itself is full of uncertainty. So it is essential to identify and analyze the software risk. Now lots of researches focus on the ERP project risk analysis, which apply a number of qualitative methods and quantitative methods. But most of methods ignore the dynamic nature of ERP project, only paying attention to predict in advance or analyze afterwards, which cannot meet the challenge of ERP project management. So Bayesian network which can provide real-time analysis is selected to analyze the ERP project risks. Through literature review and investigation of Chinese software professionals, the paper finds out 12 risk factors. In terms of an ERP project of automobile mould factory in china, revising above-mentioned risk factors, according to easy expansibility of topological structure and strong self-learning ability of Bayesian network, the paper uses prior knowledge and posterior knowledge to analyze the risk of the ERP project in requirements period. Actual application shows that Bayesian network provide an effective system method for software project risk analysis.
