Intelligent Risk Assessment in Construction: A Multi-Algorithm Machine Learning Approach for Predictive Claim Management

Sanket Patil, Sumedh Mhaske

Abstract: Construction projects frequently encounter unexpected claims resulting from cost overruns, schedule delays, and contractual disputes, creating significant financial and operational challenges for stakeholders. This research presents a comprehensive machine learning-based framework for predicting construction claims by analyzing critical risk factors including project characteristics, contract types, payment delays, design completeness, and dispute resolution mechanisms. A Random Forest Classifier was implemented to process historical project data from 45 infrastructure projects in Maharashtra, India, enabling accurate prediction of claim probabilities, cost impacts, and severity assessments. The model demonstrates exceptional performance with R? scores of 85% for time-overrun prediction and 87% for cost overrun prediction, while integrating Earned Value Management (EVM) principles to enhance predictive accuracy. The study reveals that litigation costs typically range between 50-150 million INR, with highway and expressway projects experiencing the highest schedule slippages, exceeding 20% of planned duration. By leveraging predictive analytics, this research contributes to early claim detection and proactive risk management, offering valuable insights for project planners, contractors, and decision-makers in the construction industry.

Keywords: Construction claims, Earned Value Management, Cost overrun, Time overrun