RNN-KNN FUSION FOR PREDICTIVE MODELLING OF CLIMATE INDUCED ECONOMIC DISRUPTIONS

Abstract

Climate change poses a significant threat to global economic stability, with recent studies estimating 
that unmitigated climate change could reduce global GDP by 11-14% by 2050. Sectors such as 
agriculture, insurance, and infrastructure are particularly vulnerable, with crop yields projected to 
decline by up to 25% and natural disaster damages increasing by over $300 billion annually. Despite 
this, most climate-related economic assessments remain reactive rather than predictive, relying 
heavily on manual data aggregation, static models, and domain-specific heuristics. These approaches 
lack scalability and fail to capture nonlinear climate-economic interactions. To address these 
limitations, we propose an AI-driven climate impact estimator that integrates historical climate data, 
socio-economic indicators, and industry-specific metrics to forecast the economic implications of 
climate change across sectors and geographies. The system utilizes a combination of deep learning for 
spatiotemporal climate modeling, and ensemble regression techniques for economic impact 
prediction. Additionally, it incorporates industry-adaptive feature selection and transfer learning to 
account for varying regional sensitivities and data scarcity. Our experiments on multi-sector datasets 
demonstrate strong predictive accuracy, with R² scores exceeding 0.89 for regression tasks in sectors 
such as agriculture and energy, and F1 scores above 0.85 for classifying regions by risk level. This 
approach provides a scalable, data-driven framework for policymakers and businesses to proactively 
assess and mitigate the economic risks of climate change.