Software project risk probability assessment based on dynamic Bayesian network

Abstract

Traditional Bayesian network (BN) can only have static analysis which could not reflect the impact of time factors on project risk adequately. For this reason, a software project risk probability assessment model based on dynamic Bayesian network (DBN) is proposed, which combines time series theory and Bayesian theory together to express the risk factor status change relationship between different time segments through probability and directed acyclic graph. Moreover, in the case of lack of sample data, using Leaky Noisy-or gate model to calculate the conditional probability of the nodes will come to a more objective evaluation result. Compared with the assessment results of static Bayesian network (SBN), dynamic Bayesian assessment model improves the accuracy of risk probability assessment of software projects, and provides a more scientific basis for risk control.