Conceptual Estimation of Construction Costs Using the Multistep Ahead Approach

Abstract

Providing accurate forecasts of construction costs at the conceptual phases of building projects is vital since they form an objective benchmark for the subsequent evaluation of project performance. Previous works adopted a conventional approach in which a restricted set of macro project determinants which are available in the preplanning phase was employed towards direct estimation of construction costs. Aiming to reduce the prediction error in conceptual estimates the current study adopts a novel approach from the domain of forecasting. This multistep ahead (MSA) approach relies on the idea of using several cascaded estimations to predict future values. Accordingly building element quantities were estimated as the first step. In the second step estimated quantities were combined with the existing set of inputs to achieve a higher accuracy in construction cost prediction. In order to test the hypotheses of interest 657 building projects from Germany were analyzed using linear regression and artificial neural network methods. Conclusive evidence suggests that the MSA approach significantly outperforms the prediction accuracy of the conventional practice. To the best of authors' investigation the current study is the first to offer such a cascaded estimation approach. Therefore further empirical evidence is necessary prior to generalizing applicability of the MSA approach in construction cost estimation.