• A
  • A
  • A
  • АБB
  • АБB
  • АБB
  • А
  • А
  • А
  • А
  • А
Обычная версия сайта

Препринт

Asymmetric Accuracy Metrics in Food Retail Sales Forecasting: an Empirical Comparison

This study covers the application of asymmetric accuracy metrics in the daily retail sales prediction problem. The paper is focused on the empirical validation of an accuracy metric derived from the newsvendor model. We scrutinize the accuracy metric’s advantages and describe its properties. This paper uses two main accuracy metrics: quantile-weighted and mean absolute error. We compare the economic effect of accuracy metrics for different models trained on food retail chain data. The results show that the asymmetric loss based on quantile-weighted absolute loss leads to lower business costs than the mean absolute error. The research is of interest to store managers, inventory management specialists, and logistics specialists in retail and restaurant chains. Our findings provide a better understanding of how to implement forecasting methods in order to obtain more accurate sales predictions that meet practitioner expectations.