Proactive Risk Identification in New Zealand Infrastructure Projects: A Machine Learning Approach to Classifying Project Risk

In the construction industry, proactive identification and management of challenges such as delays, budget constraints, and regulatory hurdles are essential for effective risk management and successful project delivery. Traditional methods, relying on manual analysis and expert judgment, are often time-consuming, inconsistent, and subjective. This study proposes a data-driven framework to automate the identification and classification of challenges within project descriptions using Natural Language Processing (NLP) and machine learning (ML) techniques. Utilizing the Pacificon dataset, comprising 254,923 infrastructure projects across New Zealand, the study defines challenge categories through regular expressions and implements a machine learning pipeline with TfidfVectorizer and Logistic Regression. The framework achieves a cross-validation mean accuracy of 74.6%, demonstrating robustness and reliability. Experimental results highlight strong performance in categories like regulatory hurdles and budget constraints while identifying areas for improvement, such as environmental factors. By automating challenge identification, the framework reduces reliance on manual risk assessments, enhances consistency, and enables early intervention for improved project management outcomes. This research contributes to the digital transformation of construction risk management, offering scalable, data-driven solutions aligned with industry demands for efficiency and precision in managing project complexities.