Checklist CAIO 2025 National Qualifier · Task 2
Applied Problem Solving: Loan Default Prediction
Predict whether each loan is paid back from borrower and loan features, with emphasis on detecting defaults.
The task
Part 2 of the National Qualifier is a practical problem in Google Colab. The official notebook '2025 CAIO National Qualifier Competition (Round 2)' describes a loan-default dataset of 120,000 applications (100,000 train, 20,000 test) with 12 features: annual income, debt-to-income ratio, credit score, loan amount, interest rate, gender, marital status, education level, employment status, loan purpose and grade/subgrade (A–F plus 1–5). The target loan_paid_back is 1 for repaid and 0 for defaulted, with about 80% repaid.
The notebook provides a baseline pipeline (standard scaling, one-hot encoding and a class-balanced decision tree), a validation split and code that writes submission.csv to the contestant's Google Drive folder, which is then shared with the organisers. External data is not permitted; any machine-learning approach is allowed.
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(100,000 rows with target),test.csv(20,000 rows) in the Drive data folder;test_with_target.csv(test labels) was published after the contest.- You submit
submission.csvwith columns id andloan_paid_back(probability in [0, 1] or binary 0/1) for all test ids.- Scoring
- F1 score of the default class (class 0) on 5,000 undisclosed samples randomly selected from the 20,000 test rows. The Prepare page also lists code quality as an evaluation criterion.
- Rules
- 90 minutes (Prepare page); online on the contestant's own computer.
- No external data; AI tools such as ChatGPT prohibited; only approved webpages.
- Format
- 2025 CAIO National Qualifier, 23 November 2025, 1:30–4:30 PM (per the Prepare page). Online, individual.