Demand Forecasting for Textile Products Using Statistical Analysis and Machine Learning Algorithms

Abstract

The generation of an accurate forecast model to estimate the future demand for textile products that favor decision-making around an organization's key processes is very important. The minimization of the model's uncertainty allows the generation of reliable results, which prevent the textile industry's economic commitment and improve the strategies adopted around production planning and decision making. That is why this work is focused on the demand forecasting for textile products through the application of artificial neural networks, from a statistical analysis of the time series and disaggregation in different time horizons through temporal hierarchies, to develop a more accurate forecast. With the results achieved, a comparison is made with statistical methods and machine learning algorithms, providing an environment where there is an adequate development of demand forecasting, improving accuracy and performance. Where all the variables that affect the productive environment of this sector under study are considered. Finally, as a result of the analysis, multilayer perceptron achieved better performance compared to conventional and machine learning algorithms. Featuring the best behavior and accuracy in demand forecasting of the analyzed textile products.