Demand forecasting in an automotive company: An artificial neural network approach

Abstract

This work proposes the development of two Artificial Neural Network (ANN) models for demand forecasting in the automotive industry. The networks are involved for predicting the demand of eighteen car components for a company based in the North of Italy. Statistical Package for Social Sciences (SPSS) was used as software for developing the ANNs, by setting the automatic architecture selection. The structure of the two ANN models is similar; they only differ for the partitioning of the historical data provided by the company itself respectively into training, testing and the optional holdout phases: in the first, which is the one returning the best result, data are simply assigned according to a pre-fixed percentage, while in the second a partitioning variable is introduced.