2019 CSCE Annual Conference - Laval (Greater Montreal)

2019 CSCE Annual Conference - Laval (Greater Montreal) Conference

Potential of Bayesian Networks for Forecasting the Ripple Effect of Progress Events

Return to Session

Mr. Kareem Mostafa, University of Waterloo (Presenter)
Dr. Tarek Hegazy, University of Waterloo

As part of project control, project managers need to frequently update project progress and forecast project completion time. Current methods for status analysis and forecasting rely on developing performance indices such as the well-known Schedule Performance Index (SPI) and the Estimate at Completion (EAC) of the Earned-Value (EV) analysis. These indices, however, are developed with the assumption that the remaining part of the project will follow the same trend as the latest progress trend, without regard for how current events may shape future ones or how the experienced changes and interruptions may affect future ones (i.e., the ripple effect of progress events). As such, the general EV assumptions of continued progress trend might make the forecasting results misleading. While simulation and uncertainty analysis could improve forecasting, there is still a need for more accurate forecasting methods. This paper investigates the utilization of Bayes theorem of conditional probability and the benefits of developing Bayesian networks to predict project completion time based on a certain event(s) affecting the ongoing and/or the upcoming project tasks. The paper first presents an extensive literature review of existing applications of Bayesian Analysis in various domains. To facilitate accurate forecasting, a register of possible events that trigger changes in future forecasts is identified, and includes events such as productivity loss in similar tasks, payment delays, potential site congestion, etc. The paper then discusses the potential advantages as well as the challenges of developing Bayesian models. A hypothetical seven-activity schedule to is then presented to demonstrate the proposed concept and highlight the difference between schedule updates with and without Bayesian relationships among task durations. The flexibility and timeliness of Bayes Analysis could serve to quantify potential project delays which assists in decision-making regarding proactive remedial actions on construction projects.