DATA MINING CUP 2019

20th anniversary edition: 149 Teams from 114 universities in 28 countries

The number of self-checkout stations is on the rise. This includes stationary self-checkouts, where customers take their shopping cart to a scan station and pay for their products. Secondly, there are semi-stationary self-checkouts, where customers scan their products directly and only pay at a counter. The customers either use their own smartphone for scanning or the store provides mobile scanners. You will probably have encountered this already.

This automated process helps avoid long lines and speeds up the paying process for individual customers. But how can retailers prevent the trust they have placed in customers from being abused? How can they decide which purchases to check in an effort to expose fraudsters without annoying innocent customers?

Scenario

An established food retailer has introduced a self-scanning system that allows customers to scan their items using a handheld mobile scanner while shopping.

This type of payment leaves retailers open to the risk that a certain number of customers will take advantage of this freedom to commit fraud by not scanning all of the items in their cart.

Empirical research conducted by suppliers has shown that discrepancies are found in approximately 5 % of all self-scan transactions. The research does not differentiate between actual fraudulent intent of the customer, inadvertent errors or technical problems with scanners.

Task

To minimize losses, the food retailer hopes to identify cases of fraud using targeted follow-up checks. The challenge here is to keep the number of checks as low as possible to avoid unnecessary added expense as well as to avoid putting off innocent customers due to false accusations. At the same time, however, the goal is to identify as many false scans as possible.

The objective of the participating teams is to create a model to classify the scans as fraudulent or non-fraudulent. The classification does not take into account whether the fraud was committed intentionally or inadvertently.

Download

Task
Realclass

The Winners

First Place:

Team Uni_State_Iowa_2
Iowa State University, USA

Winning team of the DATA MINING CUP 2019 from Iowa State University

Prize: 2,000.00 EUR

Second Place:

Team Uni_Geneva_1
University of Geneva, Switzerland

Second place of the DATA MINING CUP 2019 from University Geneva

Prize: 1,000.00 EUR

Third Place:

Team Uni_George_Washington_1
The George Washington University, USA

Third place of the DATA MINING CUP 2019 from The George Washington University

Prize: 500.00 EUR