Tendencias estacionales de productos farmacéuticos basadas en inteligencia artificial en la ciudad de Quevedo 2022-2024

This study examined the seasonal demand for pharmaceutical products at a pharmacy in Quevedo, Ecuador, using statistical analysis techniques and machine learning models. A dataset of 89.590 monthly sales records from November 2022 to December 2024 was employed. The workflow included data cleaning, imputation of missing records, and organization of variables for analysis. We applied seasonal-trend decomposition using locally estimated scatterplot smoothing (STL), as well as Holt–Winters, SARIMA, Random Forest, XGBoost, and long short-term memory (LSTM) neural network models. Among these, the LSTM model achieved the best performance according to the root mean square error (RMSE), followed by Holt–Winters. These techniques enabled forecasting demand for approximately 3.800 products, uncovering monthly and quarterly patterns that can guide inventory management enhancements and anticipate replenishment needs. To facilitate practical deployment, an interactive Streamlit application was developed, featuring comparative tables, charts, performance metrics, and a six-month forecasting module. It also incorporates alerts for high stock levels or potential stockouts, allowing procurement planning without requiring specialized technical expertise.

• Auteur principal: Ordóñez Guerrero, Karina Michelle (author)
• Format: masterThesis
• Langue: spa
• Publié: 2025
• Sujets: Demanda farmacéutica; Predicción de ventas; Descomposición STL; Holt-Winters
• Accès en ligne: https://repositorio.uteq.edu.ec/handle/43000/8801