Duration forecasting in resource constrained projects: A hybrid risk model combining complexity indicators with sensitivity measures

Abstract

This study combines complexity measures from the project scheduling literature and sensitivity measures from the risk analysis literature to improve project duration forecasts in resource constrained projects. A hybrid risk model is proposed incorporating project network measures, resource-related indicators, and risk sensitivity metrics. The hybrid risk model is then used for forecasting the duration of unseen projects. The study contributes to the existing literature by integrating newly proposed activity sensitivity metrics and network and resource related indicators in project forecasting. Additionally, it conducts a large-scale experiment to compare the accuracy of the hybrid risk model against benchmark methods, including Monte Carlo simulations and relevant machine learning algorithms. The results show that inclusion of resource-related variables within the hybrid risk model significantly improves the accuracy, validating recently proposed metrics. The hybrid risk model outperforms most of the benchmark methods in high-uncertainty projects, emphasizing the importance of accurately estimating the flexibility in activity start times. Furthermore, the hybrid risk model of this paper is particularly effective for parallel projects, demonstrating a better performance under various uncertainty and flexibility conditions. Finally, the results are validated using empirical project data.