Using neural network to predict performance of design-build projects in Singapore

Abstract Design-build (DB) project success may be operationalised into 11 performance metrics. 65 factors that may affect DB project success are identified. Using data from 33 DB projects, correlation analysis shows that there are several factors that affect each performance metric significantly. Artificial neural network (ANN) technique is used to construct the models to predict project performance, and these models are tested using data from five new projects. This study finds that six performance metrics can be predicted with a reasonable degree of accuracy: project intensity; construction and delivery speeds; turnover, system and equipment quality. The key variables that affect project performance may be attributed to both contractors and clients. To ensure project success, contractors should have adequate staffing level, a good track record for completion on budget, and ability in financial management and quality control. Consultants should have a high level of construction sophistication, and have handled DB projects in the past. Clients also play an important part in ensuring DB project success. They would need to have construction experience and handled DB projects in the past. In addition, they should decide on the optimal level of design completion when the budget is fixed and tenders are invited. It is recommended that owners and contractors take note of the factors identified in this study, which significantly affect DB project performance.