Modelamiento De Gastos Para La Operación De Centro De Atención Farmacéuticos
Cargando...
Fecha
Autores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Resumen
La investigación se centró en predecir los gastos de centros de atención farmacéuticos, crucial para la planificación financiera en salud. Se evaluaron varios modelos predictivos para identificar el más preciso, buscando un margen de error por debajo del 10 % de MAPE. Se utilizaron datos ajustados de gastos farmacéuticos para construir y evaluar modelos como Regresión Lineal, Regresión Ridge, Random Forest, (GBM) y modelos de ensamble. Se aplicaron técnicas de reducción dimensional como PCA y ajustes en los datos, incluyendo la filtración de valores extremos. Las métricas de desempeño evaluadas incluyeron R2, MAE y MAPE. Los modelos de ensamble, especialmente Random Forest y el modelo de Ensamble aplicados al Modelo Ambulatorio, mostraron los menores errores. La reducción dimensional con PCA no mejoró la precisión y, en algunos casos, la redujo. Aunque no se alcanzó el objetivo de un MAPE del 10 %, se logró una mejora significativa respecto a otros modelos, obteniendo un MAPE del 16 %. Se concluye que los modelos de ensamble son los más efectivos para predecir gastos farmacéuticos debido a su capacidad para manejar la complejidad de los datos. Se recomienda explorar técnicas adicionales y ajustes en los datos para mejorar la precisión de los modelos
The research focused on predicting pharmaceutical care facility expenditures, crucial for healthcare financial planning. Several predictive models were evaluated to identify the most accurate one, seeking a margin of error below 10% of MAPE. Adjusted pharmaceutical expenditure data were used to build and evaluate models such as Linear Regression, Ridge Regression, Random Forest, (GBM) and ensemble models. Dimensional reduction techniques such as PCA and data fitting, including outlier filtering, were applied. Performance metrics evaluated included R2, MAE and MAPE. Ensemble models, especially Random Forest and the Ensemble model applied to the Ambulatory Model, showed the lowest errors. Dimensional reduction with PCA did not improve accuracy and, in some cases, reduced accuracy. Although the target of a MAPE of 10 % was not achieved, a significant improvement over other models was achieved, obtaining a MAPE of 16 %. It is concluded that ensemble models are the most effective for predicting pharmaceutical expenditures due to their ability to handle the complexity of the data. It is recommended that additional techniques and adjustments to the data be explored to improve the accuracy of the models
The research focused on predicting pharmaceutical care facility expenditures, crucial for healthcare financial planning. Several predictive models were evaluated to identify the most accurate one, seeking a margin of error below 10% of MAPE. Adjusted pharmaceutical expenditure data were used to build and evaluate models such as Linear Regression, Ridge Regression, Random Forest, (GBM) and ensemble models. Dimensional reduction techniques such as PCA and data fitting, including outlier filtering, were applied. Performance metrics evaluated included R2, MAE and MAPE. Ensemble models, especially Random Forest and the Ensemble model applied to the Ambulatory Model, showed the lowest errors. Dimensional reduction with PCA did not improve accuracy and, in some cases, reduced accuracy. Although the target of a MAPE of 10 % was not achieved, a significant improvement over other models was achieved, obtaining a MAPE of 16 %. It is concluded that ensemble models are the most effective for predicting pharmaceutical expenditures due to their ability to handle the complexity of the data. It is recommended that additional techniques and adjustments to the data be explored to improve the accuracy of the models