Back to work

Risk modeling project · Case 03 · 2024

Credit Card Default Prediction

Can payment history and account behavior identify default risk without hiding minority-class errors behind a comfortable accuracy score?

A person holding a credit card while using a laptop
Role · Technical ownerPublic dataset & sourceBanking · Classification

The project

The challenge

I developed an end-to-end credit-risk analysis on 30,000 public UCI records, compared interpretable and nonlinear classifiers, tuned the decision threshold, and delivered the model through a Shiny interface.

My role

I owned data preparation, exploratory analysis, stratified sampling, feature work, logistic and Random Forest experiments, threshold evaluation, reporting, and the Shiny prediction interface.

82.4%accuracy
76.0%balanced accuracy
30kclient records

Approach

Four clear steps

  1. 01Profile
  2. 02Prepare
  3. 03Compare
  4. 04Deliver

Result

What the evaluation showed

Threshold selection improved the decision rule beyond an unquestioned 0.50 default.

Raw accuracy82.4%
No-information rate86.9%
Balanced accuracy76.0%

Outcome

What changed

The project produced a complete trail from data cleaning and exploratory analysis through modeling, threshold selection, reporting, and an interactive prediction interface.

Caveat

What this does not prove

The UCI dataset is a benchmark sample, not evidence that the model is suitable for live underwriting.
Analysis repository Shiny application
Next projectBanking & Regulatory RAG