| The project delivery system (PDS) of a construction project is one of the most important critical success factors which have great influence on project performance, an appropriate PDS could enhance the probability of project success. It has been proved that different projects suit for different PDS in various conditions, therefore the selection of PDS is crucial for a project. However, due to the fuzzy nature of PDS selection process, the model-based methods were found have difficulties in solving the problem. The instance-based method, represented by Case-Based Reasoning, solve new problems by reusing the previous experiences was thought to be suitable.The typical CBR method represents PDS selection cases by attribute-value vectors which lead to three problems, first it is insufficient to represent a PDS selection case by a set of attributes without considering their relations; then the assignment of attribute weight is not accurate and may lead to errors, and it is neither sufficient nor satisfactory to assign all the attribute numerical values. Therefore this thesis introduced relational CBR method for PDS selection purpose. This thesis introduced basic theory and concepts of Relational CBR, identified PDS selection criteria to be the attributes which represent cases and set up a hypothetical structure of the attributes. Then, this thesis gathered data from real construction projects via questionnaires to perform a confidential factor analysis which checked and confirmed the attribute structure. Then this thesis calculated local and global weight of the attributes using loading factors form CFA, and introduced the concept of structural similarity. Finally a new similarity mechanism, based on LAUD, was introduced which considered both the structural similarity and feature similarity between cases, and the Relational CBR model for PDS selection was built. At the last part of this thesis, the results of the research, the limitations of the study and future works were analyzed and discussed. |