Supply Chain Forecasting in a fast-moving global economy: Review, Limits and Future Directions
Résumé
The supply chain demand forecasting field has evolved to meet customer demand, preventing lost sales opportunities, and reducing maintenance costs. However, the large number of supply chain components (producers, vendors, warehouses, etc.) made it impossible to assess the quality of one
forecasting method on all these components. In this paper, a comprehensive literature review has been done to identify the main challenges of the supply chain forecasting. This was followed by classifying these challenges based on the type of problem presented in each paper. Furthermore, experiments have been conducted to evaluate forecast quality of the widely-used artificial intelligence models. To model the effects of recent events, such as the COVID-19 pandemic and geopolitical conflicts, various data profiles with similar patterns to those events have been used. The most used neural network architecture for forecasting has been tested against a benchmark AI approach, revealing the limitations of current forecasting techniques on less stable demands. Subsequently, several research directions have been proposed.