Data Mining Cup 2016: In the money!
The results of the Data Mining Cup 2016 were anounced on the 27th of June at the prudsys personalization summit in Berlin. Both of our solutions were awarded and the whole crew was invited to travel to Berlin in order to proudly receive the awards.
In this year the task was to predict the returns for a
particular user and product in real life data of an online shop. The first places went to the the University of California, Davis, and University of Iowa State were followed by our solution on the 3th and 4th place. This magnific results mean 500€ in our pockets, but also that we were the best European team. 76 teams participated in total.
For solving the task, we made extensive use of feature analysis, feature engineering and of the excellent xgboost algorithm for inducing boosted decision trees and combined several of those trees and a variety of other state-of-the-art classifiers to an ensemble by stacking.
Our team was composed of the following student: Kim Berninger, Simon Bohlender, Jannis Hoppe, Daniel Krause, Asamah Shad, Benjamin Tumele, Felix Zöller.