Desarrollo de un agente inteligente basado en aprendizaje automático para la mejora en la toma de decisiones estratégicas en la Dirección de Sostenibilidad de ESPOL

Abstract

The development of an intelligent agent based on machine learning to improve strategic decision-making within the Sustainability Directorate of the Escuela Superior Politécnica del Litoral addresses the critical limitation that real-time energy consumption analysis essential for meeting the Sustainable Development Goals (SDGs) and the ISO 50001 standard still requires manual handling and relies on intuitive criteria. This project highlights the need for an automated solution capable of transforming the technical data from the Personalized Recommendations for Efficient Consumption (PREC) platform into clear and actionable strategic recommendations, thereby minimizing operational workload. To this end, a conversational intelligent decision-support agent was designed and implemented, based on the ReAct architecture and the Retrieval-Augmented Generation (RAG) technique. The methodology included characterizing energy consumption patterns using Machine Learning models applied to time series data. The agent leveraged its reasoning capabilities to plan and execute code queries on the energy database, ensuring that outputs were grounded in factual data. As the main result, a functional prototype consisting of a conversational interface was developed. In conclusion, the intelligent agent significantly enhances the efficiency and traceability of the decision-making process in energy management. The system supports ESPOL’s transition toward more proactive and data-driven energy governance. Keywords: Intelligent Agent, Energy Management, ISO 50001, RAG, ReAct.