Predictive modeling for rework detection in sustainable building projects

Abstract Sustainable construction practices are implemented to ensure that buildings contribute positively to environmental protection, social well-being, and economic viability throughout their entire life cycle, from design and construction to operation and eventual decommissioning. However, construction companies have documented instances of rework in sustainable buildings, despite the advantages and potential of sustainable buildings. The systemic solution to prevent rework in sustainable buildings is to predict the occurrence of rework. This research pursues two key objectives: first, to prioritise the primary predictors of rework in sustainable buildings; second, to evaluate the most suitable machine learning algorithms for accurately modelling rework occurrences by classifying the extent of rework in the sustainable buildings. The dataset consisted of 75 responses, with 17 rework predictors. Feature scaling and normalisation were performed across the dataset to standardise the features. Six machine learning models that comprised support vector machine, Adaboost, Logistic regression, a K-nearest neighbour, neural network and random forest classifier were trained to predict the occurrence of reworks in sustainable buildings. Six feature selection techniques were used to identify the most relevant predictors. Model performance was evaluated using k-fold cross-validation to ensure generalizability. The results revealed that Random Forest and AdaBoost outperformed the other algorithms. The Adaboost achieved an accuracy of 90%, with a precision of 0.80, a recall of 0.76, and an F1-score of 0.77. The random forest model achieved a slightly higher accuracy of 90%, with a precision of 0.82, a recall of 0.81, and an F1-score of 0.81. Random Forest outperforms other models, particularly in identifying "Yes–Small" rework cases, and shows solid performance across all classes. AdaBoost also performs well, especially detecting "Yes–Very Large" cases. The research finds the degree of variation, the characteristics of the client involved, the extent of time and cost overruns experienced during project execution, adherence to performance specifications, the classification of contractors engaged, the size of the projects, and the experience of clients with the nuances of the construction sector were the main predictors of reworks. The results have important theoretical and practical ramifications for further studies on the market for sustainable buildings.