Development of four-dimensional models through automatically generated construction schedules based on building information modeling and machine learning

In construction projects, schedule preparation is generally carried out manually by experienced planners, a method that is labor-intensive, time-consuming, and prone to errors. Although various solutions have been developed, most projects still rely on manual workflows, underscoring the industry need for digitalization and automated solutions. Within the scope of this thesis, an integrated framework combining Building Information Modeling (BIM) and machine learning (ML) techniques is proposed to accelerate the schedule preparation process and reduce errors. Predecessor-successor relationships were predicted on 1,390 activities compiled from 24 projects using six different ML models, with the highest accuracy models selected through parameter optimization. These data are then used to automatically generate a schedule by executing the Dynamo codes developed in this thesis via the "BIM Auto Scheduling" add-in integrated into the Revit environment. Subsequently, activity-element matching is automated in Navisworks using the "BIM Auto Scheduling – Import CSV" add-in, enabling faster and more accurate creation of a 4D simulation. The proposed model emerges as a simple, user-friendly tool that minimizes the time loss and error risks of manual planning while providing stakeholders with a comprehensive project schedule.