Credit Loan Risk
Credit default prediction and loan risk modeling — a personal data science and ML project in active development
- Role
- ML engineer / data scientist
- Stack
- Pythonscikit-learnpandasJupyterFeature engineeringModel evaluation
Private repository, available on request.
Problem
Lenders need to estimate credit default risk before approving a loan — but raw tabular data is messy, classes are imbalanced, and it is easy to build a model that looks strong offline yet fails under real decision constraints.
Solution
Credit Loan Risk is a personal data science project building a disciplined default-prediction workflow: clean data, auditable features, baseline models, and evaluation metrics that match the problem — not just accuracy on a skewed label.
Status
In development. EDA, feature design, and the first modeling experiments are underway. This page will gain concrete metrics and artifacts as the milestones land.
Highlights
- End-to-end loan risk pipeline: ingest, clean, feature engineering, train/validate, and report
- Focus on interpretable baselines and rigorous holdout evaluation before chasing complex models
- Scoped for documentation and reproducibility first
Challenges
Outcomes
- Reproducible notebook + script pipeline for training and evaluation
- Written report tying model choices to business-facing risk tradeoffs
- Portfolio-ready walkthrough of feature design and validation discipline
Demo
The interactive demo is not published yet. The spec below is current.