Fraud detection: recognize cases of fraud at an early stage

The future of data analytics: how can you really use your insights?

In the face of increasingly sophisticated fraud attempts, how can companies effectively protect their data and detect unwanted activity at an early stage? What hidden patterns in your transactions could indicate fraudulent activity and how can they be identified automatically? Discover how state-of-the-art analytics can help minimize financial risks and make your business processes more secure.

Let's talk!

Advantage

Early warning system: Reliable detection of fraud.

Data

Transaction data

Methods

Benford analysis

Challenge

The client is a leading service provider in the finance sector, where fraudulent activities only occur in very rare cases. As it is not efficient to check all transactions manually due to the low number of fraud cases, a fraud detection method was sought in order to be able to identify suspicious cases automatically.

Goal

The customer plans to analyze the transaction data available to him in order to identify suspicious outliers with regard to fraudulent activities at an early stage. The analyses are to be carried out separately for each customer and transaction day. In order to achieve valid results, the existing data must be processed and aggregated.

Data science implementation:

Three steps to your successful data product

Learn more

Solution

Taking the customer's requirements into account, eoda analyzes the data on the basis of Benford's law. This describes a regularity in the distribution of the digit structures of numbers in empirical data sets (see diagram). The law can be observed, for example, in data sets on the population of cities, monetary amounts in accounting or natural constants. The analysis script was written in the R programming language. It provides both statistical key figures (significance values, expected values, etc.) and graphical analyses.

The latter indicate whether a review of the threshold values (overall findings) is necessary. In the second project step, the R script is integrated into the customer's existing IT landscape in close coordination with the IT department. The analysis can then be initiated either automatically via a trigger - a new transaction is stored in a defined folder - or manually by the specialist department.

Result

In collaboration with eoda, the data for the planned Benford analysis was prepared and implemented at the customer's premises. The algorithm developed decides when the assumptions of the Benford analysis are violated, meaning that a value is suspicious and fraud is suspected. The customer is informed by e-mail if the values are suspicious. An employee from the specialist department then checks the suspicious case. The data analysis is carried out automatically, but can also be started manually.

It is also possible to archive past cases in order to be able to access them if required. This project is an example of how data science can be used in business processes to create added analytical value.

Get started now:
We look forward to exchanging ideas with you.

Your expert on Data-Science-Projects:

Lutz Mastmayer
projects@eoda.de
Tel. +49 561 87948-370