Parametric Cost Estimation of Resurfacing Projects Using Ridge Regression and a Generalized Feedforward Neural Network

Reliable cost estimates during the early phases of highway projects are essential to help stakeholders make an informed decision about whether to proceed with the project. Because most highway projects are sponsored by public agencies, the significance of preliminary cost estimates to the state highway agencies cannot be overstated. However, the conventionally used ordinary least square regression modeling approach has some serious drawbacks. This study employed ridge regression and a generalized feedforward neural network, which have never been used in building cost estimation models for highway projects. There are many advantages associated with those two methods. To show these advantages with the new approaches, an ordinary least square regression model was also developed. The identical data provided by the Florida Department of Transportation were used to develop cost estimation models for resurfacing projects using these techniques. After building the optimal models, this study assessed their performances basis on the same criteria, including root mean square error and root mean absolute error. The results showed that the generalized feedforward neural network model performed better than the ridge regression model. With sufficient data, these empirical modeling approaches can also be applied to develop prediction models for other types of highway projects.

Language

  • English

Media Info

  • Media Type: Web
  • Features: References;
  • Pagination: pp 718-728
  • Monograph Title: Construction Research Congress 2016: Old and New Construction Technologies Converge in Historic San Juan

Subject/Index Terms

Filing Info

  • Accession Number: 01605889
  • Record Type: Publication
  • ISBN: 9780784479827
  • Files: TRIS, ASCE
  • Created Date: May 24 2016 3:02PM