A methodology for project risk analysis using Bayesian belief networks within a Monte Carlo simulation environment

Projects are commonly over budget and behind schedule, to some extent because uncertainties are not accounted for in cost and schedule estimates. Research and practice is now addressing this problem, often by using Monte Carlo methods to simulate the effect of variances in work package costs and durations on total cost and date of completion. However, many such project risk approaches ignore the large impact of probabilistic correlation on work package cost and duration predictions. This dissertation presents a risk analysis methodology that integrates schedule and cost uncertainties considering the effect of correlations. Current approaches deal with correlation typically by using a correlation matrix in input parameters. This is conceptually correct, but the number of correlation coefficients to be estimated grows combinatorially with the number of variables. Moreover, if historical data are unavailable, the analyst is forced to elicit values for both the variances and the correlations from expert opinion. Most experts are not trained in probability and have difficulty quantifying correlations. An alternative is the integration of Bayesian belief networks (BBN's) within an integrated cost-schedule Monte Carlo simulation (MCS) model. BBN's can be used to implicitly generate dependency among risk factors and to examine non-additive impacts. The MCS is used to model independent events, which are propagated through BBN's to assess dependent posterior probabilities of cost and time to completion. BBN's can also include qualitative considerations and project characteristics when soft evidence is acquired. The approach builds on emerging methods of systems reliability.