Danske Bank implemented a system based on AI for fraud detection

The Teradata corporation announced on October 16, 2017 development in cooperation with Think Big Analytics and implementation in Danske Bank of the platform for identification of cases of fraud in real time based on artificial intelligence technologies.

In general in the system of Danske Bank methods of machine learning are implemented and models of deep learning are at the same time developed for testing of this technology.

Danske Bank

Prerequisites

The fraud detection system which was initially used in Danske Bank was based, mainly, on the rules written manually which were actively applied in bank for a long time. Because of the record number of false operations sometimes reaching 99.5% of the total number of transactions time for conducting checks significantly increased and the corresponding expenses grew. At the same time the staff of department of fight against fraud was overloaded with work, but their actual performance remained at a low level.

Falsification of data at submission of requests for the credits — extremely important and burning issue for banks. As you know, malefactors constantly improve the equipment, using the most difficult methods for commission of the attacks therefore in their detection advanced technologies, in particular, machine learning can help — Nadeem Gulzar, the head of department of advanced analytics of Danske Bank considers. — We understand that in the near-term and long term the fraud problem will only be aggravated in connection with increase in level of digitalization in banking sector and wide dissemination of applications for mobile banking. We recognize need of use of the latest technologies capable to detect malefactors not where they are today and where they will be tomorrow. Using AI technologies we already achieved reduction of number of false identifications for 50% thanks to what a half of employees of department of fight against fraud had time for the solution of more important tasks.

Project Progress

Specialists of Think Big Analytics began work in Danske Bank with fall of 2016. Their task consisted in sharing the knowledge of with the staff of department of advanced analytics of bank how using data it is possible to derive more benefit for the company in general. Work of the joint group of experts began with creation of a basic system within the infrastructure existing in bank. Then modern models of machine learning were developed for identification of cases of fraud among millions of transactions annually, and during peak load periods — among several hundred thousands transactions per minute. For ensuring transparency and increase in the trust level, along with models of machine learning, a system includes the interpretation function explaining and providing information on actions for blocking.

From a modeling position cases of fraud meet extremely seldom — about one case from 100 thousand. The group of experts managed to select the cases which are mistakenly recognized fraud from models and to lower their number by 50%. At the same time, models are capable to reveal more cases of this fraud, having actually raised indicators of detection honor for 60%.

Project Results

The platform implemented in Danske Bank uses the methods of deep learning allowing to analyze tens of thousands of the hidden signs and to monitor millions of bank online transactions in real time for obtaining information on the real facts of fraud and the cases which are mistakenly taken for fraud.

On expectations of bank, the implemented system will allow to cut down significantly expenses on assessment of false results and to increase overall effectiveness of activity, having provided a complete investment payback in the first year of operation.

The platform of advanced analytics with a scaling option, strategy and the precise plan of actions of transition to digital technologies is necessary for all banks for implementation of technologies of a research of data in the organization — Medz Ingvar, the head of customer relations department of Think Big Analytics company noted. — As for online transactions, credit cards and payments using mobile devices, for banks it is necessary the product providing protection in real time — the modern platform on identification of cases of fraud on the basis of AI technologies developed by us together with Danske Bank processes the entering transactions less than for 300 milliseconds. So when clients buy products in supermarket, a system checks transaction in real time and at the same moment reports result. Such product will be used in all financial institutions soon.