Abstract: The track record of software industry estimates shows that the failure rate of software projects is still high,and the estimate is one of the basic reasons．The innovation of estimation method does not expect a breakthrough．Through a controlled process,you can get the desired results．Proposed a process model to guide software project to launch a series of estimates related activities．The process model consists of two parts．The first is RUP estimation process model, which gives a detailed description of how to estimate for each development and management of stage.The second is to create a graphical step-by-step process model using Bayesian network inference model,which can be effectively used to estimate analysis,communication,balance,risk prediction．The step guide is to solve the problems of the definition of estimation activities,but it is not easy to form a clear view of the estimate．The features of software estimation are suitable for modeling with Bayesian network．BN workload estimation model is abstraction of the step guides.ESFQ model is the detailed model of the trade-off relationship between the critical factors of software projects．Case studies prove the applicability of the process model.

Key words: Software estimation,Estimation process,Bayesian network