Artificial neural networks have been widely used for modeling and simulation of different problems in the construction industry, including, but not limited to, regression, clustering, and classification. They provide solutions for complex problems where other modeling methods often fail. For instance, they can capture nonlinear and complex relationships between the variables while many traditional modeling methods fail. However, they have their own limitations. They often can only be trained for a specific problem with a predetermined number of inputs and outputs. As a result, any change that requires an update in the architecture of the network cannot be automatically done and require human intervention. The recent developments in the field of artificial neural networks resulted in new concepts such as neural architecture search, reinforcement learning, and neuroevolution. These new areas can provide new methods for solving past and existing problems facing the construction industry in a more efficient, elegant, and versatile manner. One of the main contributions of the recent developments is networks that can optimize their own architecture and networks that are able to evolve and change their architecture. This paper aims to briefly review the application areas of the artificial neural networks in construction engineering and management and discuss how the recent developments in this field can be applied and provide better solutions.
Full Text (PDF)