ARTIFICIAL NEURAL NETWORK METHOD FOR PREDICTING CLOTHING SALES AT DISTRO IN LAMONGAN

Abstract

The most important indicator in business is the amount of sales. The magnitude of the level of sales affects the profits earned by the company which also affects the competition in business competition. To determine a marketing strategy, it is necessary to make accurate predictions about future market conditions, so the authors initiate forecasting using the Backpropagation Neural Network method. Beckpropagation is an algorithm that can train the network in the previous situation. The training will adjust the weights in the network as new inputs for predicting sales. The pattern used is 12-10-1, meaning that there are 12 input sales values for 12 months, 10 neurons in the hidden layer, and 1 sales output value during the following month. The results of this study were obtained at epoch 211 with an MSE value of 0.00098405, an accuracy of 79.14%, while an MSE of 0.2139.