Forecasting scope creep in Egyptian construction projects: an evaluation using Artificial Neural Network (ANN) and Random Forest models

Scope creep is a common problem in construction projects, often leading to cost overruns, delays, and compromised quality. While prior research has focused on identifying factors and proposing mitigation strategies, a significant gap remains in using advanced predictive tools to forecast scope creep and its impacts on project performance. This study addresses this gap by developing a machine learning-based model to predict scope creep in Egyptian construction projects. Data was gathered through surveys, interviews, and historical records from fifty completed construction projects. Monte Carlo Simulation was employed to enhance the dataset and by incorporating variability and uncertainties. Random Forest (RF) was used to rank influential factors, such as delays in work, poor stakeholder records, and unexpected site conditions. Meanwhile, Artificial Neural Networks (ANN) were applied to forecast cost and time overruns. The ANN model, developed using MATLAB, achieved 86% accuracy when validated through real-world case study in Egypt, demonstrating strong predictive capabilities. A Graphical User Interface (GUI) was developed to improve accessibility for construction professionals. Although the model demonstrates significant accuracy, its applicability is limited to Egyptian projects and is influenced by subjective data. This study provides a practical tool to proactively address scope creep, improve project outcomes, and bridge critical gaps in the literature.