Demand forecasting using a hybrid model based on artificial neural networks: A study case on electrical products

Purpose: This work aims to evaluate demand forecasting models determine if using exogenous factors and machine learning techniques helps improve performance compared univariate statistical models, allowing manufacturing companies manage better.Design/methodology/approach: We implemented a multivariate Auto-Regressive Moving Average with eXogenous input (ARMAX) model Neural Network-ARMAX (NN-ARMAX) hybrid for forecasting. Later, we both standard forecast the electrical products in Colombian company.Findings: The outcomes demonstrated that NN-ARMAX outperformed other two. Indeed, management improved reduction of overstock out-of-stock products.Research limitations/implications: findings conclusions this are limited sell construction industry. Moreover, experts from company provided us data also selected external based on their own experiences, i.e., might have disregarded potential factors.Practical implications: suggests neural networks including variables can accuracy, promoting approach dealing planning issues.Originality/value: demonstrate convenience proposed accuracy thus provide reliable basis its implementation supply chain electrical/construction sector companies.