Data-Driven Approach to Risk Identification for Major Transportation Projects: A Common Risk Breakdown Structure

Abstract

Identifying and evaluating risks is one of the most essential steps in risk management in construction projects. When technical and managerial complexity increases in major transportation projects, this becomes even more important. Currently, project teams are assumed to identify risks mostly based on their experience and expertise. It is a major issue that some state departments of transportation (DOT) project teams lack the risk management experience. This study proposes using a data-driven approach to unify and summarize existing risk documents to create a comprehensive risk breakdown structure (RBS). As a preliminary risk identification framework, a consolidated RBS were developed, using content analysis of public risk reports by various DOTs. Then, comparison was made between the developed RBS with 70 US transportation projects' risk registers. Natural language processing techniques, bidirectional encoder representations from transformers, were employed to calculate semantic text similarity to determine what percentage of risks are covered by generic RBS. The results showed that 70 generic risk templates cover almost 81% of the identified risks in the database of 70 major projects which is about 6000 individual risks. Project parties can use these results to discuss and identify context-specific risks as a starting point. The study also determined the interactions between risk items based on their co-occurrence using historical data. Research findings revealed the importance of considering interdependencies between risks in future studies.