Machine Learning Approaches to Enable Resource Forecasting Process of Road Tunnels Construction

Increasing demand for tunneling projects, increases attention to time and cost required for their construction. Most of parameters which are affecting on the time and cost of tunnel construction are unknown. The purpose of this paper is to provide a method to predict the construction time and cost of a road tunnelusing linear regression (LR) method. In order to train the LR method, some datasets are obtained from the historical road tunnels. To verify the feasibility of the proposed method, it has been applied to a road tunnel. All of the forecasted results have been compared with the actual results obtained during the tunnel construction and the accuracy of the predictions has been investigated. According to three statistical evaluation criteria of root mean square error (RMSE), mean absolute percentage error (MAPE) and determination of the coefficient (R2), a very high accuracy has been obtained in the prediction results. The RMSE, MAPE and R2 indices have been calculated as 0.0005 days, 0.9380637% and 0.9874 for the construction time, respectively; and 7.1194 US$,0.78891593% and 0.9873 for the construction cost, respectively.