Analítica predictiva en la nube: análisis longitudinal de la demanda de medicamentos críticos mediante Gradient Boosting en farmacias SIMAN
Loading...
Date
2026-02-01
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad Tecnológica Centroamericana UNITEC
Abstract
El presente trabajo propone un modelo de plataforma digital con analítica e inteligencia artificial para optimizar la gestión informacional de la cadena de suministro de medicamentos esenciales en Farmacias Siman, sucursal San Pedro Sula. La iniciativa surge ante la persistente problemática de quiebres de stock y la dispersión de datos de inventario y ventas, que dificultan la toma de decisiones oportunas. La investigación, de enfoque mixto y alcance descriptivo, se centró en el diseño conceptual y de información para validar la factibilidad técnica en contextos de datos limitados. Como resultados principales, se identificaron seis requisitos críticos para el Producto Mínimo Viable y se seleccionó el algoritmo XGBoost como motor predictivo, alcanzando un Error Porcentual Absoluto Medio (MAPE) del 4.01% en el horizonte óptimo de 14 días. La arquitectura empresarial se especificó conforme a TOGAF ADM, empleando una infraestructura Serverless en Google Cloud Platform (Vertex AI, BigQuery, Cloud Run y Looker Studio) que garantiza escalabilidad y bajo costo operativo. La evaluación integral del sistema obtuvo una puntuación de 89/100 (Altamente Viable), demostrando la capacidad de generar un 71.7% de alertas accionables que anticipan el desabastecimiento. La propuesta se alinea con los marcos NIST AI RMF 1.0 y NIST CSF 2.0, cumpliendo con las normativas hondureñas de salud y protección de datos. Asimismo, se fundamenta en las teorías de la Cuarta Revolución Industrial y la computación en la nube, fortaleciendo la capacidad de Farmacias Siman para responder a disrupciones globales y contribuyendo a las metas de los ODS 3 y 9 de la ONU. Los hallazgos confirman que es posible desplegar soluciones analíticas robustas en microentornos farmacéuticos, mejorando la disponibilidad continua de medicamentos esenciales mediante decisiones basadas en datos.
This work proposes a digital platform model with analytics and artificial intelligence to optimize the informational management of the essential medicines supply chain at Farmacias Siman, San Pedro Sula branch. The initiative arises from the persistent problem of stock-outs and the dispersion of inventory and sales data, which hinder timely decision-making. The research, with a mixed approach and descriptive scope, focused on the conceptual and informational design to validate technical feasibility in limited-data contexts. As main results, six critical requirements were identified for the Minimum Viable Product, and the XGBoost algorithm was selected as the predictive engine, achieving a Mean Absolute Percentage Error (MAPE) of 4.01% in the optimal horizon of 14 days. The enterprise architecture was specified according to TOGAF ADM, employing a Serverless infrastructure on Google Cloud Platform (Vertex AI, BigQuery, Cloud Run, and Looker Studio) that guarantees scalability and low operating cost. The comprehensive evaluation of the system obtained a score of 89/100 (Highly Viable), demonstrating the capability to generate 71.7% of actionable alerts that anticipate shortages. The proposal aligns with the NIST AI RMF 1.0 and NIST CSF 2.0 frameworks, complying with Honduran health and data protection regulations. Likewise, it is based on the theories of the Fourth Industrial Revolution and cloud computing, strengthening Farmacias Siman's capacity to respond to global disruptions and contributing to UN SDGs 3 and 9. The findings confirm that it is possible to deploy robust analytical solutions in pharmaceutical micro-environments, improving the continuous availability of essential medicines through data-driven decisions.
This work proposes a digital platform model with analytics and artificial intelligence to optimize the informational management of the essential medicines supply chain at Farmacias Siman, San Pedro Sula branch. The initiative arises from the persistent problem of stock-outs and the dispersion of inventory and sales data, which hinder timely decision-making. The research, with a mixed approach and descriptive scope, focused on the conceptual and informational design to validate technical feasibility in limited-data contexts. As main results, six critical requirements were identified for the Minimum Viable Product, and the XGBoost algorithm was selected as the predictive engine, achieving a Mean Absolute Percentage Error (MAPE) of 4.01% in the optimal horizon of 14 days. The enterprise architecture was specified according to TOGAF ADM, employing a Serverless infrastructure on Google Cloud Platform (Vertex AI, BigQuery, Cloud Run, and Looker Studio) that guarantees scalability and low operating cost. The comprehensive evaluation of the system obtained a score of 89/100 (Highly Viable), demonstrating the capability to generate 71.7% of actionable alerts that anticipate shortages. The proposal aligns with the NIST AI RMF 1.0 and NIST CSF 2.0 frameworks, complying with Honduran health and data protection regulations. Likewise, it is based on the theories of the Fourth Industrial Revolution and cloud computing, strengthening Farmacias Siman's capacity to respond to global disruptions and contributing to UN SDGs 3 and 9. The findings confirm that it is possible to deploy robust analytical solutions in pharmaceutical micro-environments, improving the continuous availability of essential medicines through data-driven decisions.
Keywords
Cadena de suministro farmacéutica, Inteligencia artificial, Analítica predictiva, Gestión informacional, Farmacias Siman, Honduras