Box Suite Recommendation|包装优化数据集|物流管理数据集

This data was used to select box suites for a hypothetical e-commerce retailer based on a set of 5,284 candidate boxes, stored in boxes.csv, a small set of 15,000 shipments, stored in shipments.csv and used for the actual optimization, and a large set of 150,000 shipments, stored in large_shipments.csv and used for validation. Each shipment in shipments.csv and large_shipments.csv is comprised of items selected from a set of 2,324 candidate items stored in items.csv. Each item is assumed to be a 3D rectangular carton with integral outer dimensions, so that foldable items are excluded. No two items have the same sorted outer dimensions. Moreover, none of the items have special packing constraints, such as having to be height-oriented or bottom-resting. Each shipment is comprised of 1 to 8 unique items, where each unique item has quantity 1 to 8 such that there are no more than 8 total cartons in the shipment. That is, the sum of the quantities over all the unique items in a shipment is between 1 and 8. Each candidate box occupies a row in boxes.csv comprised of a unique integral box ID and three integral inner dimensions sorted in nonincreasing order. Each candidate item occupies a row in items.csv comprised of a unique integral item ID and three integral outer dimensions sorted in nonincreasing order. Each row in shipments.csv and large_shipments.csv represents all or part of a shipment and is comprised of a unique integral shipment ID, an item's integral ID, the item's quantity, and the item's outer dimensions sorted in nonincreasing order. All the items in a shipment are assigned the same unique integral shipment ID and therefore can be grouped together based on it. For additional details about the contents of these data files and how they were used, refer to the arXiv preprint: https://arxiv.org/pdf/2004.11533.pdf