Practice Summary: Cainiao Enhances the Parcel Sorting Efficiency Through AI-Generated Delivery Zone Codes

Each package must undergo several sorting operations before reaching its destination. To optimize the sorting process, Chinese express logistics companies use a three-level delivery zone code scheme that identifies the destination sorting center, logistics outlet, and local courier for packages. In this work, we approach code generation as a multiclass classification problem in the field of machine learning. We present a lightweight transformer-based architecture developed by Cainiao that is capable of predicting delivery zone codes with high accuracy, achieving 98%–99%. Integrated into logistics management systems, this approach processes tens of millions of parcels daily for major logistics providers and reduces labor costs by 3%–5% while improving sorting efficiency.

History: This paper was refereed.

Funding: B. Yuan and W. Cui were supported by the National Natural Science Foundation of China [Grants 72301170 and 72471131] and the Startup Fund for Young Faculty at Shanghai Jiao Tong University [Grant 23X010502006].