Aplicación de técnicas de minería de datos y análisis de tendencias para el pronóstico de demanda en las MiPymes del sector textil

Abstract

With the growing expansion of the textile sector, one of the topics that has generated the most interest is the forecast of demand, since it is linked to various areas within organizations, generating up to a dependency for decision-making, however, at its own expense. Sometimes it proposes a challenge to get predictions close to reality due to the volatility generated by the market. Many researchers have been using purely statistical methods to predict this demand and, in recent decades, they have been supported by the use of artificial intelligence (Machine Learning). With the growing use of electronic platforms such as search engines, sales pages, etc., the interest by organizations in taking advantage of the information generated by users to improve their services and products has increased substantially. This work consisted of generating a demand forecasting model focused on MSMEs in the textile sector using data obtained through data mining techniques in search traffic and historical information. This model was generated through a set of phases called: work configuration, data collection, cleaning and adaptation, construction and evaluation of the forecast model, using algorithms such as linear regression, Ridge and Lasso regression, K-nearest neighborns, support vector regression, and random forest. The results showed that the Ridge and Support vector regression models provided better results in terms of the error generated in the prediction compared to traditional univariate regression statistical models.