Utilización de redes neuronales para mejorar el nivel de pronóstico de exportaciones de palta fresca, empresa Avocado Packing Company, 2020

• Autor: Campos Vásquez, Fermín Neptaly
• Formato: tesis de maestría
• Fecha de Publicación: 2022
• Institución: Universidad Nacional de Trujillo
• Repositorio: UNITRU-Tesis
• Lenguaje: español
• OAI Identifier: oai:dspace.unitru.edu.pe:20.500.14414/16927
• Enlace del recurso: https://hdl.handle.net/20.500.14414/16927
• Nivel de acceso: acceso abierto
• Materia: Redes neuronales artificiales; Volumen de ventas; Error cuadrático medio; Coeficiente de correlación

This research addressed the issue of the forecast of the volume of fresh avocado sales of the company Avocado Packing Company S.A.C. by means of artificial neural networks (RNA). The objective was to determine the level of precision of the artificial neural networks used in a forecasting model for fresh avocado exports. The methodology consisted of defining the input and output variables, such as dispatch weeks, price, per capita consumption and sales volume; which were used to establish the number of input and output neurons. Subsequently, the design of the network was carried out with the Backpropagation training algorithm and the purelin function to obtain output data. The values of hidden layer neurons and learning rate were obtained according to the lowest mean square error (MSE) value and the highest value of the correlation coefficient (R) in the training stage. In this stage, the weights and bias values were also obtained for the forecasting in the validation stage. In this stage, a MSE value of 0,0187 was obtained, lower than that obtained by the ARIMA model, of 0,0243. The statistical analysis revealed that there is no significant difference between the means of the error values of both methods, however, by presenting the neural networks method with the lowest error value, it is concluded that it is the most appropriate method for carrying out the forecast.