Abstract
This study develops and tests a machine learning (ML)-based cost forecasting model against traditional earned value management (EVM) techniques. Utilizing Python for ML implementation, the research applies algorithms to a dataset of completed projects globally, evaluating their performance with metrics like mean absolute percentage error (MAPE) and percentage error (PE). The results confirmed that ML give more accurate results than the traditional methods. Thus, the initial rate showing that XGBoost is more accurate than the traditional method using Index-2 is 88%. In 23 of 25 randomly selected projects, this algorithm was more accurate. At the middle stage, the same frequency is 92.6%; later stage, the selected criterion further confirms that the ML algorithm is more accurate than the traditional method, accounting for 75% of 21 projects out of 28. By introducing ML into project management forecasting, managers could spend less time on the technical tasks in their projects. Despite its effectiveness, the study's scope is limited by a small sample size of 110 projects and the testing of only three algorithms. Future research is suggested to expand the dataset and explore additional algorithms, including neural networks and tree-based methods, to enhance forecasting precision.