Simulation of Stochastic Activity Networks

Abstract

Stochastic Activity Networks (SANs) are used in modeling and managing projects that are characterized by uncertainty. SANs are primarily managed using Monte Carlo Sampling (MCS). The accuracy of the results obtained from MCS depends on the sample size. So far the required sample size has been determined arbitrarily and independent of the characteristics of the SAN such as the number of activities and their underlying distributions, number of paths, and the structure of the SAN. In this paper we show that the accuracy of the SANs simulation results would depend on the sample size. Contrary to existing practices, we show that such sample size must reflect the project size and structure, as well as the number of activities. We propose an optimization-based approach to determine the project variance, which in turn is used to determine the number of replications in SAN simulations.