No comments yet

Construction cost estimating with Artificial Neural Networks : Utilising ANNs for conceptual construction cost estimation

Download

- Stars (0)

17 Downloads

Owner: of CoUrsE!

Version: 1.0

Last Updated: 9-2020

DescriptionPreview

Cost estimation in the early phases of a construction project has been difficult for a very long time. Achieving a higher accuracy by using data analytics for early cost estimations would benefit construction firms. In construction management, artificial neural networks (ANNs) are an often used method for cost estimation. This research will study the possibility of using NNs to perform cost estimations for construction projects with the data available within one single firm and whether these estimations deliver a high enough level of accuracy. Theoretically, it has already been proven that cost estimations with NNs appear to be promising for construction cost estimation with highly accurate results. However, most of the studies concerning cost estimation with NNs do not result in practical implementation. Due to the supervised learning, an NN is able to generalize knowledge and learn from examples, which is useful for cost estimations, as they are mostly based on previous cases. For this study, a Feedforward backpropagation (FFBP) NN was created. The developed NN was trained on 35 data samples and tested with 15 data samples. An NN can be used for (conceptual) construction cost estimating. However, during this study, it was found that the NN did not yet achieve the desired accuracy level of ≥ 80%. The best performing NN in this study reached an estimation accuracy of 69%. It can be concluded from the results that the developed NN did not achieve the desired accuracy level. However, note that for this study the NN was developed with limited availability of data. Also, some of the data has been simulated. If the same NN was trained with a higher amount of high-quality actual project cost data, the average accuracy, as well as the accuracy of the conceptual cost estimations made with the NN estimation model, would probably increase. Despite that the desired accuracy level was not reached during the study, the study still contributes to the practical adoption and application of AI and ML in construction. The results of the study were also presented to the management and cost planners of the graduation company involved during the study. They see added value in the use of NNs for conceptual cost estimation in the future.

Post a comment