DHL Express partnered with the UTS Data Science Institute to develop a machine learning model.
We were impressed by [DSI’s] real-world application of data analytics and hands-on approach. It was clear that our partnership wouldn’t just be an academic undertaking.
Ben Somerville
Senior Manager, Customs and Regulatory Affairs for DHL Express
In Australia, exporting controlled goods across borders is a complex process that is strictly regulated by Australian Border Force (ABF).
If they get it wrong, export carriers like DHL Express face the risk of huge extra costs and even hefty fines.
But identifying problematic shipments in the millions awaiting transport from Australia can be like looking for a needle in a haystack.
“One challenge export carriers encounter is how to efficiently risk assess the commodities that shippers are exporting,” says Ben Somerville, Senior Manager of Customs & Regulatory Affairs for DHL Express.
“The process by nature is manual and relies on techniques which require human oversight.”
This challenge became the launching pad for a ground-breaking collaboration between DHL Express and the UTS Data Science Institute (DSI). The goal? To develop a machine learning model to accelerate the shipment risk assessment process.
The work involved extensive collaboration between DHL Express, DSI and ABF to identify the factors that lead to shipments being flagged for further inspection or — in the worst-case scenario — being banned entirely.
In response, the DSI team developed a proof-of-concept machine learning model that was trained on data from tens of thousands of shipments and tested against millions more.
Based on the latest examination of the DHL dataset, the algorithm can already correctly predict up to 75 per cent of the total shipments that will be held.
It can also narrow down the risk factors that increase the risk of delays.
If a shipment is considered risky, the model will identify which factors are contributing to this prediction,” says DSI researcher Dr Kun Yu, “We have applied complicated machine learning models and algorithms to resolve practical problems and make the outcomes understandable to industrial staff.
The research has been translated into a simple web portal for use by DHL staff and other industry stakeholders. In the future, portals like this will support carriers to identify high-risk shipments early in the delivery process.
This gives customers time to correct mislabelled packages, reduces hold-ups at the border, and ensures shipments get to their destinations quickly, safely and legally.
“We were impressed by DSI’s real-world application of data analytics and hands-on approach. It was clear that our partnership wouldn’t just be an academic undertaking,” says Somerville.
They were very accommodating of our myriad of requirements, including having to use our own data environments and the various security measures DHL undertakes.
And the journey is just beginning: the end goal is to deliver a software product for use by all Australian export carriers, offering significant potential for true disruption within the industry at large.