<![CDATA[Building Information Modeling (BIM) has increasingly been adopted in construction projects; however, its potential for quantitative risk prediction remains underexplored. This study aims to examine the influence of BIM-based variables on project risk severity and to develop a predictive model that explains variations in schedule delay and cost overrun. A quantitative approach was employed, using data from 42 completed construction projects that implemented 3D, 4D, and 5D BIM. Four independent variables-Geometric Complexity Index, Scheduling Density Ratio, Cost Intensity Coefficient, and Object Interdependency Level-were extracted from BIM objects and analyzed using multiple linear regression and machine learning models, including Random Forest, Support Vector Machine, and Artificial Neural Network. The results indicate that the regression model explains 68% of the variance in project risk severity, with scheduling density and cost intensity emerging as the most influential predictors. Among the machine learning models, the Artificial Neural Network achieved the highest predictive accuracy, demonstrating superior capability in capturing nonlinear relationships among BIM-derived attributes. These findings confirm that structured BIM-based metrics can serve as reliable indicators for proactive risk assessment. The study contributes to integrating BIM analytics into construction risk management frameworks and highlights the importance of data-driven decision-making in improving project performance.]]>
