Prediction of site overhead costs with the use of artificial neural network based model

Overheads, especially site overhead costs, constitute a significant component of a contractor's budget in a construction project. The estimation of site overhead costs based on traditional approach is either accurate but time consuming (in case of the use of detailed analytical methods) or fast but inaccurate (in case of the use of index methods). The aim of the research presented in this paper was to develop an alternative model which allows fast and reliable estimation of site overhead costs. The paper presents the results of the authors' work on development of a regression model, based on artificial neural networks, that enables prediction of the site overhead cost index, which used in conjunction with other cost data, allows to estimate site overhead costs. To develop the model, a database including 143 cases of completed construction projects was used. The modelling involved a number of artificial neural networks of the multilayer perceptrons type, each with varying structures, activation functions and training algorithms. The neural network selected to be the core of developed model allows the prediction of the costs' index and aids in the estimation of the site overhead costs in the early stages of a construction project with satisfactory precision.