Checklist HAIO 2025 Summer Online Qualifier · Task 1
University Admissions
English title: Egyetemi Felvételi
Predict the probability that a (synthetic) applicant is admitted to university from grades, exam results, background features and county of residence.
The task
The qualifier was a Kaggle competition. The contestant trains a model on synthetic admissions data and outputs, for every applicant in the test set, the probability of admission (target column "Felvételi Eredmény": 1 admitted, 0 not admitted). The organisers state that both the data and the task are entirely synthetic, so external data are of no use.
Each applicant is described by age, gender, grade point averages for grades 9–12, final-exam scores (history, mathematics, Hungarian language and literature, and optional computer science, biology, physics, English and German, with -1 if not taken), advanced-level flags for each subject, parents' highest education, secondary-school prestige (1–10), extracurricular activities, study habits (1–8), work experience, number of recommendations (1–6), competition participation and county of residence. A GeoJSON file gives the boundaries and centroids of the Hungarian counties and Budapest, and the starter notebook hints that the distance between the place of residence and the university (at 47.472113, 19.062236) influences admission; a haversine helper is supplied.
The starter notebook loads the data, maps the counties with folium and fits a baseline logistic regression on the school-prestige feature alone.
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(6,885 rows; 30 feature columns plus "Felvételi Eredmény" and ID),test.csv(765 rows; features and ID) and counties.geojson (county boundaries in WGS84). All three files are also published in the repository folder adatok/.- You submit
- A CSV with header "ID,Felvételi Eredmény" giving a probability between 0 and 1 for every test ID, uploaded to Kaggle and named <
received_id>_<checkpoint>.csv (e.g.HUN123_1.csv); the notebook belonging to the selected submission, named the same way, had to be submitted through a Google Form within 10 minutes after the competition closed. - Scoring
- ROC-AUC. Per the rules, if more than 80 contestants took part, the top 80 on the public leaderboard advanced; with fewer participants, or if more reached the published benchmark score, everyone exceeding the benchmark advanced, up to 100 contestants. The README and the site state that a score above the baseline was required to qualify.
- Rules
- Two hours; individual work only, no collaboration and no outside help.
- Internet use permitted only for publicly available, official sources.
- At most 20 submissions; one submission had to be selected for final scoring.
- The Kaggle team name had to be set to the personal identifier sent by e-mail (e.g. HUN123).
- Any package or framework may be used, provided the rules on the Kaggle competition page are respected.
- Format
- Summer Online Qualifier (Nyári Online Válogató) on Kaggle, spring 2025; two-hour individual round that decided entry to the national final.