在没有充足样本数据时,为有效地管理各种风险,本文提出了一个全新的研发项目风险评估模型,这个模型主要是基于Noisy-or gate 和贝叶斯网络进行评估。该模型在贝叶斯网络节点满足构成Noisy-or gate模型的前提下,利用历史数据或专家判断得到有效的网络参数,进而推断出每一项风险因素发生的概率,结合风险影响权重得到综合风险影响值,并对比得到高风险因素,为风险管理提供依据。通过与AHP方法评估结果对比表明,该模型可以准确地评估研发项目的风险,从而提高风险管理的效率。

In the case of lack of sample data, this paper proposed a new research and development (R D) project risk assessment model based on Noisy-or gate and Bayesian network for effective risk management. The model can obtain effective network parameters using historical data and expert knowledge when the Bayesian network nodes can meet the requirements of Noisy-or gate. And then we can infer the occurrence probability of each risk factor and get the integrated risk impact values combine with risk weight. So we can get the high risk factors through comparison and then provide the basis for risk control. The result contrast with AHP shows that the model can effectively assess the risk of R D projects, thereby increasing risk management efficiency.

国家自然科学(71072123)；中央高校基本科研业务费项目(FRF-BR-12-020)；北京科技大学研究生教育发展基金