LaDe: The First Comprehensive Last-mile Delivery Dataset from Industry

TL;DR

LaDe addresses the critical need for publicly available, real-world last-mile delivery data by introducing the first industry-scale dataset with pick-up and delivery records across multiple Chinese cities. Comprising LaDe-P (pick-up) and LaDe-D (delivery), it offers large-scale, comprehensive, and diverse features suitable for route prediction, ETA estimation, and spatio-temporal graph forecasting, with baseline benchmarks and public code. The dataset enables multi-task research and cross-city analysis, while also highlighting challenges such as data missingness and privacy perturbations. By releasing LaDe on HuggingFace, the work aims to accelerate advances in logistics optimization, spatio-temporal data mining, and related AI applications in the last-mile domain.

Abstract

Real-world last-mile delivery datasets are crucial for research in logistics, supply chain management, and spatio-temporal data mining. Despite a plethora of algorithms developed to date, no widely accepted, publicly available last-mile delivery dataset exists to support research in this field. In this paper, we introduce \texttt{LaDe}, the first publicly available last-mile delivery dataset with millions of packages from the industry. LaDe has three unique characteristics: (1) Large-scale. It involves 10,677k packages of 21k couriers over 6 months of real-world operation. (2) Comprehensive information. It offers original package information, such as its location and time requirements, as well as task-event information, which records when and where the courier is while events such as task-accept and task-finish events happen. (3) Diversity. The dataset includes data from various scenarios, including package pick-up and delivery, and from multiple cities, each with its unique spatio-temporal patterns due to their distinct characteristics such as populations. We verify LaDe on three tasks by running several classical baseline models per task. We believe that the large-scale, comprehensive, diverse feature of LaDe can offer unparalleled opportunities to researchers in the supply chain community, data mining community, and beyond. The dataset homepage is publicly available at https://huggingface.co/datasets/Cainiao-AI/LaDe.

Figures (5)

• Figure 1: Overview of LaDe from last-mile delivery (better viewed in color), which includes two sub-datasets: LaDe-P from package pick-up process (i.e., couriers pick up packages from sender customers and return the depot) and LaDe-D from delivery process (i.e., couriers deliver packages from the depot to receiver customers).
• Figure 2: Region-level and AOI-level data.
• Figure 3: Spatial and temporal distribution of data in Shanghai of LaDe-P.
• Figure 4: Diversity of cities. We select two cities, Hangzhou and Jilin, as an example to reveal their different spatio-temporal distributions. (a) The time distribution of packages in a day; (b) The ETA distribution of packages; (c) The distribution of the average distance between two consecutive packages in a courier's route. A significant difference is observed in the above illustration.
• Figure 5: Illustration of three real-world applications. (a): Route prediction predicts the future pick-up route of a courier. (b): ETA prediction estimates the courier's arrival time for picking up or delivering packages. (c): STG forecasting predicts the future package number in given regions/AOIs.