engleski

Comparative Analysis of Linear Regression and Soft Computing Methods for Estimating Highways Construction Time and Cost in the Republic of Croatia

Construction of road structures, including highways, on the Republic of Croatia territory often suffers from the planned time and cost deviations, whereby overruns are particularly unacceptable. In terms of costs, overestimated costs can also cause problems because it is more demanding to ensure more investment money. Therefore, the main research goal is to investigate relations between realized and contracted time, also realized and contracted costs of highway construction by applying predictive modelling – linear regression and some soft computing methods. In addition, the goal is to investigate the possibility of establishing the appropriate model for estimating the duration and cost of similar projects by using simple linear regression analysis, neural networks (NNs) and support vector machine (SVM), using the software DTREG. By comparing the accuracy of the obtained models using mean absolute percentage error (MAPE) and the coefficient of determination (R2) as standard estimators of the model accuracy, the optimal models have been selected, and the possibility and conditions of their application in practice have been analyzed. Furthermore, the differences, advantages, and disadvantages in applying linear regression and soft computing methods in solving this and similar engineering problems were explored, primarily in time and cost planning.

roads ; highways ; time ; cost ; linear regression ; neural networks ; support vector machine

nije evidentirano