AI in Construction Management: Preparedness and potential. A case study on implementing a predictive machine learning framework for construction project scheduling

Abstract

Persistent challenges related to project delays continue to plague the construction industry, an industry often characterized as outdated, low-productivity, and unpre dictable. These challenges are amplified by the complexity of infrastructure projects, external pressures, and the historically slow adoption of digital technologies. De spite generating large volumes of data, the industry struggles with inconsistent data collection and effective utilization. To address these limitations and underline the importance of robust data management, this thesis explores the integration of ML based predictive models to improve decision-making in project management. In col laboration with NCC, the complex Ingelkärr–Stenkullen transmission line project served as a case study. A hybrid forecasting model was developed, combining Monte Carlo simulations with a neural network-based ML approach. The Monte Carlo simulations generate a wide range of potential project completion timelines, incorporating variations in task durations and task-specific characteristics. These simulated outcomes serve as the foundational training data for the neural network. A key technical contribution of this work lies in the model’s dynamic weekly updates with real-world progress data, enabling adaptive learning. Task dependencies were processed using GPU acceleration, and an attention mechanism allowed the neural network to capture task interactions, enhancing predictive accuracy. Interviews with NCC and Svenska Kraftnät project managers and engineers informed the model’s user interface, ensuring transparency and improved decision-making. Results showed high predictive accuracy (R2 = 0.92), which improved over time, highlighting the value of combining data-driven methods with traditional management strategies. Ultimately, this thesis demonstrates the critical need for next generation planning systems in construction, focusing on intelligence, adaptability, and transparency. The proposed framework shows strong potential to transform industry practices by significantly improving risk forecasting, optimizing resource management, and in creasing responsiveness to uncertainty, thereby offering a pathway to more efficient and resilient project management in construction