ABSTRACT
The inherently complex nature of risks interdependencies in construction projects coupled with incomplete data records during projects development often results to inaccurate assessments. This paper showcases the use of neural networks for risk assessment in construction projects. A detailed literature review identifies the different types and training methods of neural networks as well as the respective tools applicable to construction projects risk management. Based on these findings, the paper presents the development of a specific neural network that partially assesses occupational risk in a construction engineering project. The proposed neural network is trained with metadata from previous risks assessments. The modeling of the network is realized through two software tools, in order to identify potential difficulties in the modeling process as well as potential deviations in the assessments’ outputs. The main conclusion is that neural networks are reliable for conducting risks assessments that realistically integrate risks interdependencies in complex problems.
