Investigating the possible regression functions for modelling delays in infrastructure projects in South Asia

The aim of this paper is to contribute to the studies on schedule delays by analyzing the characteristics of delays, investigating regression functions for modeling the correlation between time overruns and influencing factors, and identifying the causes of delays in infrastructure projects across South Asia. A dataset of 138 completed projects in the region is collected for multiple regression function analysis to identify the best-fit function for modeling the complex pattern between time overruns and influencing factors. To detect the causes of overruns, a content analysis enabled by NVivo software is utilized. A significant contribution of this study is introducing a novel non-parametric regression function as an alternative to the traditional statistical regression functions for modeling the correlation between time overruns and other variables. The mean value of time overruns in infrastructure projects within South Asia is 86.66 %. The main root causes of delays are land acquisition and resettlement, procurement delay, contractor delay, and design revision. Overall, machine learning algorithmic techniques such as random forest regression provide more efficient and flexible models and deepen understanding of complex patterns between time overruns and other variables. The practical application of this study is to serve as a reference for planning future projects.