| This dissertation describes in detail the development of the survey to assess the value being added by the owner based on recommended roles and responsibilities. The survey yielded 73 projects from 45 different companies. By first defining project success from the perspective of each of the project participants, the research was able to determine the key owner functions and characteristics that contribute to project success. Through the use of descriptive and inferential statistical analysis, several regression models were developed to assist the owner in determining the correct level of involvement on the project. Two paradigms are investigated. One is classical statistical technique---multiple regression, the other is a modern artificial intelligence technique---neural networks. This study shows that both traditional regression analysis and artificial neural networks deliver beneficial project success analysis. However, regression analysis provides the baseline for the neural network method. Each has unique advantages and disadvantages, and hence each is appropriate for distinct types of problems. The combination of statistics and artificial neural network yield the best results. |
