Abstracts – Browse Results

Search or browse again.

Click on the titles below to expand the information about each abstract.
Viewing 1 results ...

Chan, T W (2007) Improving the estimation of project overheads in construction companies in Hong Kong, Unpublished PhD Thesis, Department of Civil and Building Engineering, Loughborough University.

  • Type: Thesis
  • Keywords: artificial neural network; estimating; Hong Kong; overheads; prototype development; quantity surveying
  • ISBN/ISSN:
  • URL: https://hdl.handle.net/2134/34843
  • Abstract:
    Project overheads cover the site cost of administrating a project as a whole, rather than a particular work section. Estimation of these items is one of the routine tasks of all parties including the contractors and project owners. Nevertheless, our understanding of this subject mainly lies on the theoretical level due to the limited empirical study in the past. More importantly, estimation of project overheads demands a lot of expertise but still exhibits a high risk of inaccuracy. Therefore, the aims of this research study are to explore the estimation and expenses of project overheads in practice and to devise an efficient model for project overheads estimation. In the data collection process, surveys and interviews were conducted with large contractors in Hong Kong. A comprehensive review on the expense pattern, estimating method and accuracy of project overheads was this developed upon the empirical base. To improve the estimating accuracy and efficiency, and estimating model utilising artificial intelligence was designed. The model was an artificial neural network (ANN) model adopting the group method of data handling (GMDH) algorithm. Input variables were determined by an opinion survey, followed by exploratory factor analysis to extract the principal factors for modelling. The model was trained with 63 project cases collected from contractors in Hong Kong and validated by another eight cases which were not being used in the training process before. Satisfactory training and testing results were obtained, together with an identification of five significant variables affecting project overheads. The model had undergone further validations including eight rounds of cross validation; comparison with linear multiple regression; and comparison with other ANN architectures like multi-layered feed-forward network and general regression network. All the results evidenced that the proposed GMDH model was a reliable tool, producing accurate predictions on project overheads. To affirm the applicability of the model, a focus group of seven senior quantity surveyors was conducted. Members of the group conceded that the model was efficient and accurate; and worth further development into a tailored model for the individual company.