Análisis de riesgos financieros y empresariales derivados de la escasez hídrica, producto de la sequía hidrológica prolongada, en Ecuador desde el año 2023: Un enfoque basado en la sostenibilidad empresarial y predicción de amenazas mediante la aplicación de Inteligencia Artificial y Machine Learning en sectores estratégicos directamente afectados

Abstract

The 2023-2024 water crisis in Ecuador imposed critical financial risks on productive sectors. This study develops a predictive and prescriptive Early Warning System to assess corporate vulnerability in Azuay through a hybrid methodology integrating Data Envelopment Analysis, Natural Language Processing, and Machine Learning. Salience analysis identified the Government as the entity of maximum urgency, with a score of 3.0000, evidencing absolute institutional dependence. Using the DEA model applied to a longitudinal panel of 460 DecisionMaking Units from the agricultural, manufacturing, and energy sectors, a critical disparity in technical efficiency was detected. Industries with high asset rigidity showed a systemic setback, reflected in a 58.7% decline in the Malmquist Productivity Index. For the predictive phase, an Extreme Gradient Boosting algorithm was trained that achieved excellent accuracy in cross-validation. The analysis showed that the Environmental Exposure Factor and dynamic productivity account for over 71% of predictive power, outperforming traditional ratios such as static liquidity. Finally, generative artificial intelligence integration via document retrieval allowed processing the regulatory environment to prescribe mitigation tactics. The study concludes that corporate sustainability in the face of extreme climate shocks requires abandoning static accounting and adopting dynamic models that measure technological efficiency and exogenous exposure.