Application of Artificial Neural Networks for the Prediction of Cashflows in Public Road Works

Abstract

An attempt was made to predict the cashflows of public road projects using artificial neural networks. In the context of the development of prediction models, an introduction to the financial flows, to the Greek legislation that defines them, and finally to artificial intelligence was made. Also, a literature review concerning the application of artificial intelligence in the construction industry was carried out. Neural networks were then applied based on 37 public road projects. The methodology highlighted three models for the prediction of cashflows for public road projects: a statistical exponential regression model, an artificial neural network model, and finally a hybrid model that combined the two previous ones. The hybrid model had the lowest mean absolute prediction error, followed by the model using only artificial neural networks, and lastly the statistical regression model. Finally, the conclusions of the study, the limitations that existed, suggestions for the application of the model, and ideas for future research are presented.