Development of a support vector machine-based predictive model for bored pile productivity in residential construction projects in Iraq

Accurate prediction of construction productivity remains a critical challenge in the civil engineering sector, particularly for deep foundation works such as bored piles in residential projects. This study proposes a data-driven predictive model based on Support Vector Machine (SVM) to estimate the productivity of bored piles in high-rise residential construction projects in Iraq. Real-world data were collected from the Iraq Gate Residential Complex and used to train and validate the model. Key influencing factors included pile geometry, soil type, equipment specifications, crew size, and working hours. The model achieved a mean prediction accuracy of 99.89% and a correlation coefficient (R) of 97.02%, demonstrating superior performance over conventional estimation methods. These findings highlight the practical value of machine learning approaches in enhancing resource planning and decision-making during early project phases. The proposed SVM-based model can support contractors and engineers in forecasting performance outcomes and minimizing scheduling uncertainties in similar construction settings.