Climate-Adjusted Machine Learning-Driven (Re)Insurance Pricing Using Future Projections and Disaster Frequency-Average Annual Loss Dynamics

There is a growing demand in reinsurance for parametric modeling frameworks that are not only fast and computationally efficient, but also capable of incorporating real-world, forward-looking scenarios based on observable and projected risk drivers. In response, this study proposes an integrated, climate-adjusted framework for natural catastrophe (NatCat) pricing that combines Average Annual Loss (AAL), machine learning-based disaster frequency modeling, growth-rate attribution, and reinsurance pricing metrics. Using country-level hazard and exposure data, Random Forest models are employed to jointly estimate disaster frequencies from observed AALs and, conversely, to infer AALs from modeled disaster frequencies, thereby ensuring internal consistency across pricing components. Growth rates are quantified at both aggregate and hazard-specific levels and projected under climate scenarios for 2030, 2050, and 2100. The proposed framework enables a forward-looking assessment of climate-driven risk evolution and supports risk-based pricing decisions with direct practical applicability for insurers, reinsurers, and public risk pools engaged in underwriting, capital management, and climate-resilient risk transfer mechanisms. The contribution of this study lies in the integration of machine learning-based frequency estimation, climate-adjusted growth-rate attribution, and reinsurance pricing within a single, internally consistent NatCat pricing framework, rather than in the development of new hazard or climate models.