An Overview of Performance Measurement for Demand Forecasting Based on Artificial Neural Networks

Robus, Steffen and Kőműves, Zsolt Sándor and Walter, Virág (2022) An Overview of Performance Measurement for Demand Forecasting Based on Artificial Neural Networks. REGIONAL AND BUSINESS STUDIES, 14 (2). pp. 5-20. ISSN 2061-2311 (print); 2732-2726 (online)

Preview

Text 01_Robus.pdf Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (620kB) | Preview

Abstract

Demand forecasting is an essential task to match supply and demand. From a supplier’s view, demand forecasting is important to optimize supply chains and thus maximize profits. The ever-increasing availability of data that can be used as input factors for predictive models allows more and more sophistication for diverse forecasting tasks in the context of demand forecasting. On the one hand, increasingly complex models have been used for demand forecasting over the last years, from simple exponential smoothing methods and ARIMA models up to complex, hybrid (deep) artificial neural networks. On the other hand, little attention is paid to the methods that evaluate the forecasting performance of these models, which are essential for the selection from among potential forecasting models. In this article, we aim to answer the question of what are the most favourable measurements in recent literature on applied neural network demand forecasting for supply chain management. To this end, we analyzed 193 relevant publications in which demand forecasting was applied using artificial neural networks. We found that in artificial neural network demand forecasting used to evaluate forecasting performance, Mean Absolute Percentage Error, Root Mean Squared Error, Mean Squared Error and Mean Absolute Error are by far the most popular methods. Furthermore, we found that when forecasting performance measurements are combined, the most common combination is the combination of Mean Absolute Error, the Root Mean Squared Error and the Mean Absolute Error.

Actions (login required)

Edit Item