基于贝叶斯网络的项目组合风险度量

摘要

The project portfolio risk cannot be assessed by the linear addition of single project risks be-cause of the interaction effects.This method combines expert knowledge and K2 algorithm to obtain the Bayesian network of project portfolio risk, and calculates the conditional probability table of nodes in the network by measuring the impact of interaction effects on project risks,so as to realize the Bayesian network inference of project portfolio.To generate an ordered variable input required by K2 algorithm, the mutual information between project risks is calculated, which is combined with conditional independent test to ob-tain the ordered project nodes.Finally, the practicability and effectiveness are proved through applying this method to a new high-tech enterprise.%项目组合的交互效应特性使得项目组合风险不能通过单个项目风险的线性叠加获得。基于贝叶斯网络建模提出了一种项目组合风险度量的新方法。该方法通过将专家知识与K2算法相结合，求得项目组合风险的贝叶斯网络结构，并通过度量交互效应对项目风险的影响计算网络中每个节点的条件概率表，实现项目组合风险的贝叶斯网络推理。为了得到K2算法所需的有序节点输入，计算项目风险间的互信息，并基于互信息与条件独立检验求得项目节点的顺序。最后通过一个高新技术企业项目组合的应用实例说明该方法的实用性和有效性。