A Machine Learning-Driven Framework for Minimizing Construction Delays

Technology, Khulna, Bangladesh nahinahmedchowdhury29@gmail.com Construction delays remain a persistent challenge in construction project management, often causing budget overruns and schedule disruptions. This study proposes a machine learning (ML) based framework for predicting and mitigating delays, integrated with 4D Building Information Modeling (BIM) to improve schedule precision for C-shaped residential structures in Autodesk Revit. A literature-driven synthetic dataset of 500 construction projects was assembled, capturing activity durations, delays, and task dependencies. The scheduling data were refined in Primavera P6 and modeled in Autodesk Navisworks to enable 4D simulations. Four machine learning techniques, random forest (RF), support vector regression (SVR), gradient boosting trees (GBT), and extreme gradient boosting (XGBoost), were evaluated using mean absolute error (MAE), root mean square error (RMSE), and the coefficient of determination (R²). XGBoost achieved the strongest performance, reducing prediction error by 31% compared to the SVR baseline. When applied to a simulated C-shaped residential case, the framework identified geometry complexity and substructure phase activities as the leading delay drivers using Shapley additive explanations (SHAP), enabling project teams to allocate schedule buffers proactively. This reduced total predicted delay by 94.4% and schedule deviation from 73.8% to 4.1%. These risk rankings can be embedded directly into Primavera P6 buffer allocation workflows, giving practitioners in construction markets prone to monsoon delays, such as Bangladesh, a quantifiable basis at the activity level for scheduling geometrically complex residential developments. Impact of Individual Attributes on Construction Readiness Parameters for Resuming Abandoned Housing Projects: A Multivariate Regression Analysis N. Aishah Abdul-Rahman1,2*, Rahimi A. Rahman1,3, Ahmad Rizal Alias1, Syafizal Shahruddin4, Md.