Predicting Highway Construction Costs: Comparison of the Performance of Random Forest, Neural Network and Support Vector Machine Models

Inaccurate cost estimates have substantial effects on the final cost of construction projects and erode profits. Cost estimation at conceptual phase is a challenge as inadequate information is available. For this purpose, approaches for cost estimation have been explored thoroughly, however they are not employed extensively in practice. The main goal of this paper is to comparing the performance of various models in predicting the cost of construction projects at early conceptual phase in the project development. In this study, on the basis of the actual project data, three modeling algorithms such as random forest, support vector machine and artificial neural networks are used to forecast the construction cost of Ethiopian highway projects. The three models were then compared based on the outcomes of prediction and root mean square error. The findings revealed that random forest outperforms neural network and support vector machine in realizing better prediction accuracy. Based on root mean square error, the random forest cost model provides 18.8% and 23.4% more accurate result than neural network and support vector machine models respectively. It is anticipated that a more reliable cost estimation model could be designed in the early project phases by using a random forest regression technique in the development of a highway construction cost estimation model. In conclusion, the practitioners in the highway construction industry can make sound financial decisions at the early phases of the project development in Ethiopia.