Modelo de pronóstico de ventas de vehículos livianos usando redes neuronales artificiales

Abstract

Accuracy in sales prediction is vital in any type of industry as it can improve the quality of business strategies or decision making. Specifically, in the automotive industry it plays an increasingly important role due to the growing competition registered in the market, to the long development and production times and to its relationship with the economic variables. This study compares the performance of the classical forecasting methods against artificial neural networks (ANN) when forecasting light vehicle sales of the Neohyundai company, internal and external variables characteristic of the sector are considered. First, a preprocessing step was carried out to ensure the quality of the input data to the ANN, in this stage the normalization and selection of the most influential variables through Lasso regression is included, then Network architecture and learning parameters are established. The results show that ANN exceed traditional prediction methods, this means that the demand for vehicles that are durable consumer goods can be predicted by using different variables. Furthermore, when determining the relative importance of the input variables, it was found that the economic variables contribute significantly to the forecast.