Predicting Cost Overruns in Utility Streetworks

Organisation || BAM Energie en Water
Project Type || M.Sc. Thesis
Candidate || Rick Potkamp

Over the past six months, I have researched whether cost overruns in utility projects can be predicted based on enriched (historical) project data. This issue is becoming increasingly relevant as the number of projects in the utility sector is expected to rise due to, among other factors, the energy transition. As a contractor, it is crucial to have a clear understanding of the risks associated with cost overruns.

Stages in Project aimed at Predicting Cost Overruns in Streetworks

My research consisted of three phases. In the first phase, I identified influential factors affecting cost overruns in utility projects through literature review and expert interviews. My goal was to determine key features that have a significant impact, even if this project data is not yet actively collected or used.

For my client BAM Energie & Water, I developed data-driven models to predict cost overruns. The goal was to create a supportive, predictive model that helps estimators assess whether a project will experience a cost overrun (binary classification). The results indicate that a Random Forest model is capable of doing so, with an accuracy of 63.7%, a recall of 63.9%, a precision of 62.4%, and an F1-score of 63.7%. Based on discussions with the client, these scores have been deemed sufficient for a supportive model. Further research can be conducted to improve performance.