Детальная информация

The construction industry is increasingly affected by cost overruns, schedule delays, resource inefficiencies, and uncertainty in project execution. Traditional construction risk management methods, such as expert judgment, probability-impact matrices, the Analytic Hierarchy Process (AHP), and Monte Carlo simulation, provide useful support but often remain subjective, static, and limited in their ability to process large multidimensional project data. This master’s dissertation develops a data-driven integrated risk intelligence framework for predictive and sustainable construction project management using supervised machine learning techniques. The research is based on a structured dataset of 500 building construction project records, including project characteristics, planned cost, planned duration, building type, structural system, resource indicators, environmental conditions, and complexity-related parameters. The methodology includes data preprocessing, categorical encoding, scaling, feature engineering, model development, validation, interpretation, and practical implementation. Several machine learning models are examined for predicting actual cost, actual duration, cost overrun percentage, schedule overrun percentage, risk level, risk score, and scenario classification. Ensemble-based methods, particularly Extra Trees and Random Forest, are emphasized because of their ability to capture nonlinear relationships in construction project data. The developed framework not only provides numerical predictions but also supports decision-making through risk classification, cost and schedule risk scores, similar project case analysis, and explanation of possible risk causes. The selected model is implemented in a bilingual Streamlit-based web application, allowing users to enter project data, receive predictive outputs, and download a PDF report. The results show that predictive risk intelligence can support early identification of project risks, improve cost and schedule control, assist resource planning, and contribute to more sustainable construction management by reducing rework, waste, and inefficient resource use. The study demonstrates that integrating machine learning with practical decision-support tools can transform construction risk management from a reactive process into a proactive, data-driven, and sustainability-oriented approach.