Neural networks applications in estimating construction costs

Abstract

This thesis deals with the potential application of neural networks technology to construction cost estimating problems. This is done by developing neural networks applications for a number of case studies constructed from the historical cost data of actual construction projects.

Parameter-based cost estimating applications, which require the application of analysis and prediction techniques to the cost data of a given estimating problem, were chosen as the major field of investigating the implementation of neural networks in this thesis. The objective of this thesis is to investigate whether or not neural network computing technology should be considered as a viable alternative in cost estimating applications by comparing it with conventional parameter-based analysis tools or predictive methodologies currently used to estimate construction costs. Both methodologies, parametric estimating and neural networks, use a parameter-based approach in modeling cost. However, the computational techniques used by the two methodologies to analyze cost data and produce results are significantly different.

Four case studies were the subject of comparison. The four case studies were compiled from the records of two construction companies and focus mainly on two areas: (1) Industrial projects and (2) Bridge construction.