Innovative Out-of-Stock Prediction System Based on Data History Knowledge Deep Learning Processing

Research and development efforts in the field of commercial applications have invested strategic interest design intelligent systems that correctly handle out-of-stock events. An event refers to a scenario which such customers do not availability products they want buy. This generates important economic damage producer store. Addressing problem is currently great as it would allow limiting damages deriving from these Furthermore, era online commerce (e-commerce), significantly limit events show considerable impact field. For reasons, authors proposed solution based on deep learning for predicting residual stock amount product analysis specific visual–commercial data well seasonality. By means combined pipeline embedding convolutional architecture boosted with self-attention mechanism downstream temporal network, will be able predict remaining particular commodity. integrating interpreting climate/seasonal information, customers’ behavior data, full history dynamics sales, possible estimate certain and, therefore, define purchase orders efficiently. accurate prediction stocks allows an efficient trade order policy results significant reduction The experimental confirmed effectiveness approach accuracy (in products) greater than 90%.