FreshRetailNet-LT: A Stockout-Annotated Censored Demand Dataset for Latent Demand Recovery and Forecasting in Fresh Retail
Researchers have introduced FreshRetailNet-50K, a novel dataset designed to improve demand forecasting for perishable retail goods. This dataset addresses the challenge of censored sales data, which occurs when stockouts prevent the observation of true customer demand. By providing detailed hourly sales data with stockout annotations from nearly 900 stores, FreshRetailNet-50K enables more accurate latent demand reconstruction and subsequent forecasting, showing a significant improvement in prediction accuracy and bias reduction. AI
IMPACT Enables more accurate demand imputation and perishable inventory optimization by addressing limitations in retail AI.