Abstract: A risk evaluation model in software project investment based on Bayesian Networks（BNs） is presented in this paper.The BNs parameter learning is applied to the modeling process based on sample data set，so that the BNs is more accordant with the project feature in the software project investment phase.In addition，the validity of the parameter learning is validated with algorithm precision and convergence.Practice proves that the risk evaluation model can provide the accurate risk information for decision-makers in the process of software project investment.

摘要： 提出了一种基于贝叶斯网络的软件项目投资风险评价模型。在建模过程中以样本数据集为基础进行贝叶斯网络参数学习，从而在软件项目的投资阶段建立更加符合实际项目特征的贝叶斯网络。同时，从算法精度以及算法收敛性这两个方面验证该参数学习过程的有效性。经实践检验，在软件项目投资过程中该风险评价模型能够向决策者提供准确的投资风险信息。