| Research was undertaken to develop a model to predict cost overruns on land-based, North American construction projects. A belief network model was chosen to meet this goal. First, a catalogue of all risks to construction cost was compiled. These risk variables were ranked from most dependent variable to most independent variable to aid in model development. Relationships between each of the risk variables on the list were determined by surveying experts in the field, Thirdly, conditional probabilities for each of the relationships in the model were calculated, again with the aid of expert surveys. Finally, the model was calibrated using data from actual completed projects.;The user may set values of each of the variables in the model, and read the likelihood of a range of cost overruns for each of the Equipment, Labour and Material Costs on a construction project. Two completed projects were tested, and the differences between actual and predicted cost overrun were found to be within 10%. |
