GDP Data Mining Challenge 2026 :
Erasmus teams successfully enhance AI models

In the last week of May 2026, 25 students from 7 universities met at Fulda University of Applied Sciences for a data mining hackathon. Around 50% of the participants were female - an atypical distribution for the field of computer science. After two weeks of online courses, they travelled to Germany for the face-to-face part of a so-called Blended Intensive Program (BIP), supported by the Erasmus+ programme. Such programs are significantly shorter than a full semester and enable students with occupations, family commitments or other restrictions to gain international experience.

The course began with a brainstorming session on feature engineering. Afterwards, both standard prompts and sophisticated, agent-based prompts were used to expand this list. The more advanced prompts were able to significantly increase the variety of features, but human brainstorming was still able to add value in diverse teams. Once the features were implemented, the five competing teams managed the model selection and tuning. The modern XGBoost algorithm performed significantly better than the alternative algorithms and was fine-tuned accordingly. While two teams with a focus on fewer but more relevant features achieved slightly better results, all teams worked successfully and gained additional knowledge in data mining. One of the students concluded: "I have attended several AI courses, but this one gave me the full picture of what it means to use data mining in practice."

Professor Frank Klingert from Fulda University of Applied Sciences led the data mining hackathon and also based the course on his own Erasmus experience, which he gained 20 years ago at Leiden University. He was supported by Floor Weijman from Saxion University of Applied Sciences and Tatiana Blömer from Fulda University of Applied Sciences, who helped the students to form their groups and reflect on their work with the help of a silent retrospective. The students from Babeș-Bolyai University, Jamk University of Applied Sciences, Linnaeus University, New Bulgarian University, Polytechnic Institute of Setúbal, St. Pölten University of Applied Sciences and Fulda University of Applied Sciences returned home with new insights into data mining and many positive impressions of collaboration in European teams.