AI-ENHANCED MARKOV CHAIN MODELS FOR LONG-TERM SUSTAINABILITY PREDICTION IN INFRASTRUCTURE PROJECTS

Abstract

Infrastructure projects in developing economies face escalating socio-economic and environmental risks that threaten long-term sustainability. Traditional risk management approaches, predominantly static and deterministic, prove inadequate for capturing the dynamic and probabilistic nature of risks in complex environments. This study addresses this critical gap by integrating artificial intelligence (AI) classification algorithms with Markov Chain models to enhance predictive accuracy and support sustainable project outcomes. Employing a mixed-methods design with 442 stakeholders across eight Niger Delta infrastructure projects, we developed AI-enhanced transition probability matrices using supervised machine learning techniques. The Complex Tree classifier achieved 100% accuracy in stakeholder role prediction, while regression analysis demonstrated that AI-Markov integration explained 78% of variance in sustainability performance (β=0.88, p<0.001). Comparative ANOVA testing revealed that AI-enhanced Markov Chains significantly outperformed traditional Bayesian Networks (Cohen's d=0.68, p=0.021) and Bow Tie analysis (d=0.86, p=0.009) in tracking temporal risk evolution. The integrated framework enabled dynamic recalibration of risk forecasts, with steady-state convergence occurring within 10-15 assessment periods across all case studies. These findings demonstrate that hybrid AI-probabilistic models offer substantial advantages for sustainable infrastructure delivery in volatile environments, though implementation barriers including data infrastructure gaps and stakeholder skepticism remain. The study contributes a robust methodological framework for adaptive risk management while providing actionable insights for policy reform and capacity building in developing regions.