<ul><li>Accurate demand estimation is crucial for the retail industry to guide inventory and pricing decisions for perishable items.</li><li>FreshRetailNet-50K is introduced as a significant benchmark dataset for censored demand estimation, containing 50,000 store-product time series with detailed hourly sales data and annotations for stockout events.</li><li>The dataset offers unique temporal granularity and contextual covariates like promotional discounts and weather, enabling innovative research in demand modeling and forecasting.</li><li>The two-stage demand modeling approach demonstrated with FreshRetailNet-50K shows improved prediction accuracy and reduction in systematic demand underestimation, paving the way for advancements in demand imputation and retail analytics.</li></ul>

FreshRetailNet-50K: A Stockout-Annotated Censored Demand Dataset for Latent Demand Recovery and Forecasting in Fresh Retail

Discover more