A Framework for the Characterization of Aviation Construction Projects: The Case of UAE

Abstract

This article contributes to the existing literature by modeling and automating the learning process from previous aviation construction projects (ACPs) using artificial intelligence tools, where it will be easier to characterize aviation construction projects and identify the specifications of different aspects of the projects throughout their entire life cycle. An artificial intelligence (AI) framework is proposed for the categorization of aviation construction projects using different machine-learning (ML) methods with a focus on the UAE as a source of data. Airport construction projects have been seen to share a good deal of similar attributes, which should simplify the decision-making process regarding layouts, design, equipment, labor, budget, complexity, etc. However, the gap in reality is that the huge and scattered sources of data, project specifications, characteristics, and the knowledge from past projects are not utilized in an automated way that could simplify the navigation through projects for better future decision-making. The utilization of AI/ML tools is expected to be useful here in order to reduce the revisions of design and construction rework by classifying the projects and the elements that managers need to consider. The planning, design, and construction of new projects can be improved by identifying the attributes of past projects and categorizing them according to similarities, differences, and complexities. Specifically speaking, a framework of hierarchical clustering and neural networks is integrated together to form the classification model. Upon implementing hierarchical classification and neural networks, it was found that neural networks could demonstrate remarkable classification results; the error in classification was minimal in most of the cases. The advantage of such classification is to help decision-makers utilize best practice from the groups of previous projects, which were classified using both hierarchical and neural networks models. With this classification, rework can be minimized, overhead costs may be reduced, and past best practices can be utilized.