Estimation Model of Construction Project Procurement Performance Based on Organization Experiences

• Main Authors: Chih-Sheng Hsiao, 蕭志勝
• Other Authors: Li-Chung Chao
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/91375041222507575981

博士 === 國立高雄第一科技大學 === 工程科技研究所 === 101 === Determining the form of procurement contract and the accompanying method for contractor selection constitutes an important decision of the owner of a construction project, which is often influenced by organizational factors. The objective of this research is to develop an organizational experience or organizational memory based approach to predicting procurement performance and to achieve it, multiple regression models and data-clustering-based fuzzy models are proposed for predicting construction project procurement performance. Taiwan Power Company with a continuous demand for contracting out construction was studied as an example and its 96 substation projects completed in recent years through three procurement routes, i.e. traditional, design-build, and turnkey, were collected. Besides collecting the quantitative data on each project including floor area, cost and speed, a questionnaire survey was done for obtaining the assessment data for qualitative factors on the environment, project, owner/consultant, and contractor, as well as levels of satisfaction about procurement performance; these two types of data were combined for use for model development. First, the results of regression analysis show that 20 of the 39 models developed achieve Adj R2 values higher than 0.6 and floor area is the variable that impacts on construction speed most significantly. The testing errors of the three best models for each procurement route against nine reserved cases average between ?{19.6% and 7.6%. Next, for developing the fuzzy models, through a factor analysis, an initial set of 48 variables were first reduced to nine inputs, while eight performance metrics were used as model outputs. Then, using the fuzzy clustering method along with hybrid training, Sugeno-type fuzzy inference systems establishing the input-output relationships were obtained. Tests show that zeroth-order systems outperform both first-order systems and regression models in prediction accuracy, and can predict project performance under each route reasonably well, while sensitivity analyses confirm their robustness. As this research used performance data for projects of the same type undertaken by the same firm, the prediction models developed can represent organization experiences and have organizational learning effects, which can be used as an aid in procurement decision.