Search, find, click – and off to the shopping basket with the product. In an ideal world, that is how online dealers imagine the surfing and purchasing patterns of shop customers. The reality, however, is different. Not every product that is viewed and placed in the shopping basket actually ends up with the purchaser. Many transactions are interrupted for a variety of reasons before they ever reach the binding order stage.
This year, the realtime specialist and initiator of the DATA MINING CUP called on students from all over the world to test their data mining know-how and develop a model that would make it possible to predict the actual orders in an online shop. The classic data mining task consisted of using historical and real anonymous shop data from sessions and which contained information, whether or not the session ultimately resulted in an order for the shopping basket or not, to develop a model that could predict orders in further sessions in the same shop. The second task was to implement an agent that can predict the probability of an order during a session based on each individual transaction.
99 teams from 77 universities and 24 countries participated in this student competition.
|Solution Task 1|
|Solution Task 2|
Task 1 First Place:
Team 2 from Dortmund University
Task 1 Second Place:
Team 1 from Dortmund University
Task 1 Third Place:
Team 1 from the Karlsruhe Institute of Technology
The team placement for the second task was equally international. Here, first prize went to students from the Budapest University of Technology and Economics. Second and third place were occupied by students from the Anhalt University of Applied Sciences.