Modelo predictivo de las ventas de productos de primera necesidad en el sector comercial basado en técnicas de Machine Learning

Abstract

The present work proposes to develop a predictive model of the sales of essential products in the commercial sector based on machine learning techniques, using convolutional neural network, CNN, models, with which predictive analysis can be achieved. Information on sales operations corresponding to the years 2018 to 2021 was collected. The approach used is quantitative-qualitative, mixed, with an experimental design. The methodology that was adapted to the study was Cross - Industry Standard Process for Data Mining or CRISP - DM, which is a methodology oriented to data mining, which allows the levels of tasks to be reviewed hierarchically, from the most general to the most specific, in phases that go from knowing the business to the deployment of the results obtained. These results evoked in 256 combinations. To measure the best predictive performance, it was considered to evaluate 3 models according to the filters in relation to the number of layers, as well as the size of the kernel, of which model A presented the best performance of the results in the MSE metrics of 0.000300 and MAE of 0.008900 with the parameters of 64 filters, 5 kernels and 1,000 epochs.