Machine Learning-Based Prediction of Construction Project Delays in Public Building Projects in Nigeria

Construction project delays remain one of the major challenges affecting the successful delivery of public infrastructure in Nigeria, resulting in significant cost overruns, schedule extensions, and reduced project performance. This study developed and evaluated a machine learning framework for predicting delays in public building projects using historical project data from 1,200 completed projects executed between 2015 and 2022. Twenty-three project characteristics, including contract sum, project duration, payment delay, material price increase, contractor experience, project complexity, and design modifications, were used as predictor variables. The dataset was preprocessed through missing value imputation, outlier treatment, feature engineering, and standardization before model development. Six supervised machine learning algorithms comprising Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, Artificial Neural Network, and Extreme Gradient Boosting (XGBoost) were trained and evaluated using five-fold cross-validation. Results showed that 67.2% of the projects experienced schedule delays, with an average delay duration of 182.4 days. Correlation and feature importance analyses identified payment delay, project complexity, and material price escalation as the most influential predictors of project delay.Comparative evaluation indicated that the XGBoost model achieved the highest predictive performance with an accuracy of 96.67%, precision of 96.34%, recall of 96.08%, F1-score of 96.21%, and ROC-AUC of 0.989, outperforming the Artificial Neural Network (95.42%), Random Forest (93.75%), Support Vector Machine (91.25%), Decision Tree (87.08%), and Logistic Regression (84.58%). The optimized model correctly classified 232 out of 240 testing observations and demonstrated excellent robustness with a cross-validation standard deviation of only 0.34%. The developed framework provides an interpretable and highly reliable decision-support tool capable of identifying high risk projects before significant schedule overruns occur, thereby supporting proactive project management, efficient resource allocation, and improved delivery of public building projects in Nigeria.