ING uses AI in areas like transaction monitoring and customer experience, focusing on fraud detection and transaction categorization.
Fraudsters use phishing emails to steal login credentials, gain control of an account, and initiate unauthorized transactions that appear legitimate.
A hybrid approach combines a rule-based system for initial checks with a machine learning model for anomaly detection based on customer behavior.
Transactions are categorized differently based on whether they are card payments, which use merchant codes, or non-card payments, which rely on machine learning models analyzing transaction descriptions.
Machine learning models struggle with transaction descriptions in multiple languages, as performance degrades for minority languages with less training data.
A custom translation table is built by identifying frequently occurring keywords in the majority language and mapping them to their equivalents in minority languages, improving model accuracy.
29:37Stefan Donsa & Lukas Alber
24:49George Proorocu
35:16Ankit Patel
26:34Markus Hacker & Tabitha Kleine
29:28Kai Grunwitz, Klaus Bürg & Tomislav Tipurić
28:29Taqi Jaffri, Kapil Gupta & Farooq Sheikh and Tomislav Tipurić
29:30Kilian Kluge & Isabel Bär
29:34Dr. Saskia Meier-Andrae, Stefan Bär, Daniel Gebler & Keno Dreßel
Jobs that call for the skills explored in this talk.
Serviceware SE