Cost overrun prediction for building construction projects in UAE using artificial neural networks

Abstract

Construction projects are unique and dynamic in nature; as such, they evolve over their lifecycle and are affected by a number factors both internal and external to the project which ultimately reflects on cost, time, or quality. Cost overruns have been globally reported in the majority of construction projects, to the point that it is argued cost overruns are inherent to the industry. Nonetheless, the magnitude of cost overruns, risks, and factors causing overruns varies across geographical locations and projects due to a wide array of variables. This variability becomes especially challenging for estimators to account for when a project's details are not yet fully defined, typically during the early stages of a project's life cycle, which coincides with budget allocation. The objective of this research is to develop a cost overrun prediction model to be used in the pre-construction stage, tailored to the UAE construction industry to help decision makers in financial and investment planning using Artificial Neural Networks (ANN). This objective is achieved through examining the research available pertaining to cost overrun through a literature review including different modelling methods. Next, a two-step survey is distributed to experienced professionals within the UAE construction industry. The first survey collects background data and assesses the magnitude of the cost overrun experienced on the last project completed by each respondent. The cost overrun ranged from - 5% to 15% with an average of 5.81%. The second survey collects completed local building projects' records. The records are then used to develop ANN models with bootstrapping and ensemble modelling to predict cost overruns. The models predicted cost overrun within 10% error for 50% of the data set with a correlation coefficient (R2) 0.94 of and Mean Squared Error (MSE) of 0.97. The key contribution of the research is bridging the gap in literature with respect to predicting construction cost overrun in the UAE construction industry. This is achieved through assessing the magnitude of the cost overrun experienced locally for building projects and developing a reliable cost overrun prediction model. Cost overrun models can aid project stakeholders in predicting the magnitude of cost overrun a project will likely experience based on local historical records. The prediction can be implemented during budget allocation, contingency allocation, and project financing and investment decisions and analysis.