Statistical cost modelling for preliminary stage cost estimation of infrastructure projects

Abstract

Reliable and accurate cost estimates are essential to construction projects. They are even more critical in infrastructure projects as they require more time, cost, and public constraints. Therefore, a better cost model is required for infrastructure projects. An extensive literature review was carried out to identify various statistical modelling techniques and models, as well as models developed using these techniques. The literature identified seven statistical modelling techniques. They are; regression analysis, Monte-Carlo simulation, support vector machine, case-based reasoning, reference class forecasting, artificial neural networks, and fuzzy logic. These techniques were all used in various cost models developed for construction projects. According to the analysis of results, neural networks and support vector machine-based models displayed better performance in their cost estimation models. However, it was found that combining several techniques into a hybrid model, for example, the neuro-fuzzy hybrid, can significantly increase these results. Thus, the reliability and accuracy of the current estimation process can be improved with these techniques. Finally, the techniques identified as having better performance can be used to develop a cost estimation model for the preliminary stage. This is because these techniques perform well even though the availability of information is lower. The results of this research are limited to the seven identified techniques and the literature used in the review.