Diseño de un modelo computacional basado en técnicas de minería de datos para el pronóstico de la demanda de productos farmacéuticos

• Autores: Cabanillas Martínez, Elmer Jhosban, Martínez Reyes, Paul Fray
• Formato: tesis de grado
• Fecha de Publicación: 2014
• Institución: Universidad Nacional de Trujillo
• Repositorio: UNITRU-Tesis
• Lenguaje: español
• OAI Identifier: oai:dspace.unitru.edu.pe:20.500.14414/11296
• Enlace del recurso: https://hdl.handle.net/20.500.14414/11296
• Nivel de acceso: acceso abierto
• Materia: Modelo computacional; Productos farmacéuticos

Survival in the highly competitive business today requires an accurate view of demand to implement production plans, inventory, distribution and purchase within business; the pharmaceutical industry is no exception, as the effects of seasonality, promotions, price changes, advertising, products with low or high motion and outliers generally affect the determination of the same. In this context, above forecast demand among its consequences the drug excess inventory, obsolescence or expiration, and on the other hand, below the forecast demand has resulted in the loss of sales and a possible increase in costs . As mentioned, the theme focuses on the development of a computational model, which the techniques of artificial neural networks and genetic algorithms for forecasting demand for products used._x000D_ In the present work an analysis and comparison of computational models and then the data mining techniques is performed, which I decided to use the most appropriate for the development of computer modeling techniques (Which used a multilayer perceptron neural network with backpropagation algorithm optimized with genetic algorithms) with the traditional model (only used multilayer Perceptron neural network with backpropagation algorithm) for forecasting the demand for pharmaceuticals.._x000D_ The accuracies and errors obtained by each of them are: our proposed model has 98.55% accuracy and 1.45% error, the traditional model have 89.0589% accuracy and 10.9410% error._x000D_ This leads to the conclusion that neural networks can be optimized using genetic algorithms, in order to get better results for different problems are proposed