SMART DELAY PREDICTION: SUPERVISED MACHINE LEARNING SOLUTIONS FOR CONSTRUCTION PROJECTS

• Published in: Journal of mechanics of continua and mathematical sciences Vol. 20; no. 6; pp. 154 - 167
• Main Authors: Sahu, Pramodini, Bera, Dillip Kumar, Parhi, Pravat Kumar, Kandpal, Meenakshi
• Format: Journal Article
• Language: English
• Published: Institute of Mechanics of Continua and Mathematical Sciences 01.06.2025
• Subjects: adaboost; construction delay; decision tree; gaussian naïve bayes; gradient boosting; logistic regression; random forest; xgboost
• DOI: 10.26782/jmcms.2025.06.00010
• ISSN: 0973-8975

With construction project delays being a key factor influencing their financial sinews, issues related to contract law, thus rendering resources incapable, remain a contemporary issue all over the globe. Conventional techniques for predicting delays often do not deliver concrete predictions due to the multiplicity and dynamic character of construction tasks. In the study discussed here, different machine learning (ML) algorithms were investigated to foresee construction delays, and these include Gaussian Naïve Bayes, Adaboost, Logistic Regression, Gradient Boosting (GB), Random Forest (RF), Decision Tree (DT) and Extreme Gradient Boosting (XGBoost). These models were measured for performance using various metrics such as accuracy, precision, recall, and F1 score to assess their validity in real-life situations. The results indicate that the use of ensemble learning techniques such as Random Forest (RF), and XGBoost scores higher than others, thus exhibiting more accuracy and better predictive capacity. These can relate to convoluted relationships in construction data, which makes them suitable for yet another application in project risk management. In contrast, simpler models like Adaboost and Gaussian Naïve Bayes, despite being interpretable, held lesser predictive accuracy and hence were less qualified for construction delay forecasting. The study highlights the potential of ML-driven predictive models in aiding project management timely by enabling timely identification of prospective delays, thereby allowing for proactive decision-making and project control. Hence, hurdles like quality of data, interpretability of models, and integration with real-time project management systems ought to be surmounted for wide-scale adoption in the industry. Future studies ought to develop hybrid ML models that fit in explainable AI techniques and real-time data on construction applications to assure predictive accuracy and usability in practice. The findings show that ensemble-based delay prediction models have the potential to reduce the uncertainties related to projects, control delays, schedule resources efficiently, and ultimately improve the infrastructure project efficiency in costing and timely project completion.