A comparative study of regression analysis, neural networks and case-based reasoning for early range cost estimation of mass housing projects

Thesis Type: Postgraduate

Institution Of The Thesis: Middle East Technical University, Faculty of Engineering, Department of Civil Engineering, Turkey

Approval Date: 2010

Thesis Language: English

Student: Hüseyin Karancı

Supervisor: RİFAT SÖNMEZ


Construction cost estimating is essential for all of the stakeholders of a construction project from the beginning stage to the end. At early stages of a construction project, the design information and scope definition are very limited, hence; during conceptual (early) cost estimation, achieving high accuracy is very difficult. The level of uncertainty included in the cost estimations should be emphasized for making correct decisions throughout the dynamic stages of construction project management process, especially during early stages. By using range estimating, the level of uncertainties can be identified in cost estimations. This study represents integrations of parametric and probabilistic cost estimation techniques in a comparative base. Combinations of regression analysis, neural networks, case – based reasoning and bootstrap method are proposed for the conceptual (early) range cost estimations of mass housing projects. Practical methods for early range cost estimation of mass housing projects are provided for construction project management professionals. The methods are applied using bid offers of a Turkish contractor given for 41 mass housing projects. The owner of all projects is Housing Development Administration of Turkey (TOKI). The mass housing projects of TOKI are generally a mix of apartment blocks, social, health and educational facilities, and some projects may also have mosques. Results of the three different approaches are compared for predictive accuracy and predictive variability, and suggestions for early range cost estimation of construction projects are made.