Neural networks for modelling the final target cost of water projects

Producing reasonably accurate cost estimates at the planning stage of a project important for the subsequent success of the project. The estimator has to be able to make judgement on the cost influence of a number of factors including site conditions, procurement, risks, price changes, likely scope changes or type of contract. This can shroud the estimation process in uncertainty, which has often resulted in project cost overruns. The knowledge acquisition, generalization and forecasting capabilities of Artificial Neural Networks (ANN) are explored in this pilot study to build final cost estimation models that incorporate the cost effect of some of the factors mentioned above. Data was collected on ninety-eight water-related construction projects completed in Scotland between 2007-2011. Separate cost models were developed for normalized target cost and log of target costs. Variable transformation and weight decay regularization were then explored to improve the final model’s performance. As a prototype of a wider research, the final model’s performance was very satisfactory, demonstrating ANN ability to capture the interactions between the predictor variables and final cost. Ten input variables, all readily available or measurable at the planning stages for the project, were used within a Multilayer Perceptron Architecture and a Quasi-Newton training algorithm.