Author(s): Erik Johansson

Abstract: The construction industry faces persistent challenges in efficiently managing project schedules due to the dynamic nature of activities and uncertainties such as delays and resource constraints. Traditional scheduling methods, including the Critical Path Method (CPM), often fall short in adapting to real-time project disruptions. This study aimed to explore the integration of machine learning (ML) models into construction project scheduling, focusing on their potential to enhance predictive analytics and dynamic decision-making. The primary objectives were to evaluate the predictive accuracy of ML models, identify suitable approaches for task clustering, and assess the effectiveness of reinforcement learning (RL) for adaptive scheduling.A combination of supervised ML models, including Decision Tree (DT), Random Forest (RF), and Gradient Boosting Machine (GBM), along with RL algorithms, was utilized. Historical project datasets were pre-processed, analyzed, and divided into training and testing sets. Performance metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R-squared (R²) were applied to evaluate the models. K-Means clustering was employed to categorize tasks based on durations and resources, while RL algorithms were tested for real-time scheduling adaptability.The results showed that GBM achieved the highest predictive accuracy, with an R² of 0.89, and significantly outperformed traditional methods. RL-based dynamic scheduling reduced delays by 47% and cost overruns by 58%, while clustering analysis provided actionable insights for resource prioritization. Statistical analyses confirmed the significance of ML performance improvements over traditional methods.This study concludes that ML models offer transformative potential for construction scheduling by enhancing accuracy, adaptability, and efficiency. Practical recommendations include developing centralized data repositories, integrating ML with Building Information Modeling (BIM), and promoting training programs. Future research should focus on real-time data integration, scalable ML models, and cross-industry applications to maximize these benefits.

DOI: 10.22271/27078302.2024.v5.i2a.51

Pages: 28-33 | Views: 124 | Downloads: 61

Download Full Article: Click Here