🕵️♂️ Credit Default Risk Intelligence
Overview
Kaggle Home Credit Default Risk dataset: 1,000 loan applications with bureau history, credit card balances, POS transactions, and installment payments.
Data Structure
| Metric | Value |
|---|---|
| Loan Applications | 1,000 |
| Default Rate | 32% |
| Tables Joined | Bureau, credit card, POS, installments |
| Onboarded By | LLM pipeline (Opus 4-6, 52 rules) |
Causal Reasoning Model
Traces why borrowers default through multi-table causal chains:
- Weak Credit Scores: external bureau deterioration signals
- Revolving Credit Abuse: credit card overextension patterns
- Financial Stress: installment payment delays and POS delinquency
- Demographic Risk Factors: age, income, employment type compounding
Rich causal chains connect observations across tables to explain individual default decisions.
Sample Questions
- "Why did this borrower default despite appearing low-risk?"
- "What signs of financial stress preceded the default?"
- "How did credit card overextension contribute?"
Try It
Ask Gem Logic: Why did this borrower default despite appearing low-risk?
Start Session →