Cost estimation in road construction using artificial neural network

Road construction projects on the territory of the Republic of Croatia are characterized by the overrun of planned costs. The experience of the contractor on previous road projects is an important element that can help to prevent errors and increase the chances of success in similar future projects. Data on construction costs collected from past projects can be used to estimate costs at different stages of the project life cycle through artificial neural networks. In this paper, artificial neural networks (MLP, GRNN, RBFNN) for estimating road construction costs are modeled. During the modeling, the database of roads constructed on the territory of the Republic of Croatia was used. Comparison of performance of neural networks has shown that the GRNN has obtained the best accuracy with MAPE of 13% and coefficient of determination of 0.9595. The neural network has proven to be a promising approach to use in the initial design phase when there is usually a limited or incomplete set of data for cost analysis, and this method could yield much more accurate results and the estimation error could be reduced.