Checklist Benin selection 2025 Selection test (Kaggle) · Task 1
Binary Classification on Synthetic Tabular Data
Train a binary classifier on anonymised synthetic tabular data and submit predicted probabilities for the test set.
The task
The competition's short description, in French, asks candidates to train a binary classification model and to submit the predictions of their model on the test data ('Les candidats devront entrainer un modèle de classification binaire et soumettre les prédictions de leur modèle sur les données de tests').
The Overview, Evaluation, Data and Rules pages are unchanged copies of Kaggle's Tabular Playground Series – October 2021, from which the competition was cloned: the data are synthetic, generated by a CTGAN trained on real-world data on the biological response of molecules; the features are anonymised and are a mix of scaled continuous and binary columns.
For each id in the test set, the probability of the binary target must be predicted.
Abridged by SOTA from the official materials. The official statement has the exact rules, and it wins wherever this summary differs.
At a glance
- You get
train.csv(feature columns and target);test.csv;sample_submission.csv.- You submit
- CSV with header id,target giving a probability for every test id.
- Scoring
- Area under the ROC curve between the predicted probability and the observed target. Kaggle setting: the public leaderboard uses 20% of the test data.
- Rules
- Teams of up to 3 (Kaggle setting and copied rules text).
- Kaggle setting: 25 submissions per day (the copied rules text still states 5); up to 2 final submissions.
- Format
- Single online session on Kaggle on 31 May 2025 (about 2 h 24 min between enabling and deadline); 18 teams made 35 submissions (Kaggle metadata).