Home

Loan Default Prediction — DHDL Ltd.

What you'll see here

The business problem, the scoring rule, and what a good result looks like.

Training loans
93,174
Loans to predict
39,933
Default rate
23.75%
Our CV accuracy
89.6%
leaderboard top 88.28%
Context

DHDL Ltd., an Indian non-bank lender, is losing profit to rising loan defaults. It wants to know which borrower and loan characteristics drive default, and to predict default for every loan in the test set.

Metric: accuracy = (TP + TN) ÷ all loans. Predicting "no default" for everyone already scores 76.25%, so a model has to beat that bar by a clear margin.

Submission: a CSV with exactly 39,933 rows plus a header, columns ID,default.

Explore the data