Simulación de pronósticos de ventas en la empresa IMPACTEX mediante redes neuronales

Abstract

In the present research work, the simulation of the sales forecast was carried out in the IMPACTEX company using Artificial Neural Networks. For which the free Python software was used with the help of the Tensorflow and Keras libraries, the data used was historical data obtained from the IMPACTEX company for the years 2008-2019, from which a sales record was obtained by product code. (127 products in total), to carry out the forecast an ABC analysis was carried out that allowed determining the products with the highest demand, obtaining as a result the product BH1060 followed by the product BH7010 located in group A, while in group B the product C585 and group C was made up of product 1112.6. The structure for each neural network was obtained after varying the epoch parameters, the number of layers of the neural network and the number of neurons, the number of hidden layers coinciding in three for the four products used, the number of neurons per layer and epoch. varies depending on the error response obtained in the simulation. The selected parameters generated an error of 2.60 % for product type A (BH1060), for product type A (BH1070) 3.64 %, for product type B (C585) 3.02 % and a 3.27 % error for product C (1112.6)