Predicting High-Impact Research in the Construction Engineering and Management Domain Using Computational Machine Learning

Construction engineering and management (CEM) domain covers multiple evolving sub-disciplines and directions. To understand those directions, numerous research efforts have conducted scientometric analysis of publications to explore their significance and impact. Despite their valuable insights, such research endeavors have been limited to exploratory analyses, whereas no previous attempt has been conducted to predict the future impact of research in CEM. This study fills this knowledge gap by performing a predictive analysis of citation metrics of publications in CEM using machine learning (ML). The developed model was trained on a dataset of 93,868 publications using an XGBoost algorithm. Testing showed that the model could predict high-impact research after 5 years from publication with an accuracy of 80.71%. Further analysis of the predicted model results identified "project planning and design," "organizational issues," and "information technologies, robotics, and automation" as the top forecasted CEM research subdisciplines that are anticipated to attract high citation counts. Ultimately, this research contributes to the CEM body of knowledge with a framework that can predict future citation trends, which can help guide researchers and stakeholders in academia identify research and funding opportunities.