Analysis of The Application of Machine Learning in Predicting The Risk of Construction Project Delays

Abstract

Construction project delays are one of the main challenges that impact costs, schedules, and economic benefits. This study aims to analyze the application of Machine Learning  algorithms in predicting the risk of delay in construction projects, especially in the Balikpapan Refinery Development Master Plan (RDMP) project. The project's historical data for the 2020–2022 period was used to build a prediction model using  the Support Vector Machine (SVM), Random Forest (RF), and Random Forest hybrid model optimized using Genetic Algorithm (RF-GA). The analysis process includes data cleansing, variable transformation, descriptive analysis, modeling, parameter optimization, and model performance evaluation using Accuracy, Precision, Recall, F1-score, and AUC metrics. The results of the descriptive analysis showed that the Construction  discipline was the largest contributor to the cumulative deviation, while  the Procurement and Engineering disciplines were relatively stable. Cumulative Dev variables, current period deviations (This Period Dev), and Previous Dev are proven to be the main factors causing project delays. Evaluation of the model's performance showed that RF-GA had the highest accuracy of 99.34%, with a Precision, Recall, and F1-score of 0.99 each and an AUC of 1,000, outperforming SVM and RF without optimization. The RF-GA model is also able to effectively predict project delay risks in the 2024–2025 period, with the predominance of delays in the construction discipline  and consistent seasonal patterns. The findings of this study have significant managerial implications, namely the importance of monitoring time deviations from the early stages of the project, the focus of control on field activities,