Selective Classification as a means to create Business Value: Insights from the Logistic Sector

Selective classification

The logistics industry faces unique challenges with data uncertainty and complexity, such as classifying Harmonized System (HS) codes and predicting shipment importers. Current AI models often produce predictions even in cases of uncertainty, leading to potential costly errors. Xi Chen’s research focuses on selective classification, enabling AI systems to 'reject' uncertain samples and pass them to human experts instead of risking mistakes.

Confident predictions

Xi Chen demonstrates in his research how selective classification enhances decision-making in logistics by focusing on confident predictions and rejecting uncertain ones. This approach automates processes like Harmonized System (HS) code classification and importer prediction, improving productivity and accuracy while reducing delays. This can save millions of Euro's for DHL Express per year, and it inspires cross-industry applications, particularly in fields like healthcare and finance, where decision reliability is critical. Recommendations include integrating selective models into operational workflows and exploring their adaptability for broader industrial use.

PhD defence Xi Chen

Xi Chen(Jiang Xi, China, 1992 ) has been a data scientist at DHL since 2019. He graduated in 2014 from NanChang Aviation University with a major in Mechanical automation and design, and in 2019, he completed a Master’s in Data science and decision making at the Maastricht University. On Thursday, Dec 19 at 4:00pm, he will defend his dissertation titled 'Selective Classification in Logistics' at Faculty of Computer Science of the Open Universiteit in Heerlen.

Promoters are Prof. dr. Tanja E. J. Vos (Open Universiteit & Technical University of Valencia) and Dr. Stefano Bromuri (Open Universiteit). The defense can be attended live at the Open Universiteit in Heerlen and followed online via ou.nl/live.