From the analysis of software project performance evaluation indicator system, owing to its incompleteness, excessive simple evaluation model, and informal approach, using the statistic analysis theory, an indicator system about software organizations status and software projects characteristics is given. The connotation of software project performance is defined through literature research. Then, the author develops a software project performance evaluation model based on Fuzzy Neural Network using a new network topology structure, which precisely and efficiently resolves the problem of evaluation model’s connection right weights. The empirical research indicates that the model may effectively evaluate the software project performance and cognize project risk factors, provide risk avoiding measures, better project performance, and give decisionmaking support information.