Skip to content

Latest commit

 

History

History
14 lines (9 loc) · 828 Bytes

File metadata and controls

14 lines (9 loc) · 828 Bytes

Credit Risk Modeling with Expert LLM Reporting

This project demonstrates an end-to-end workflow for credit risk classification using three robust machine learning models:

  • Random Forest
  • XGBoost
  • Deep Learning

The primary goal is to maximize recall, ensuring high-risk customers are not misclassified as low-risk. The workflow addresses class imbalance using SMOTE and provides automated, expert-level interpretation of results in an executive report generated using a local LLM (Mistral via Ollama).

Data

Source: Kaggle - Credit Risk Customers

Description: Contains demographic, financial, and credit history data for customers. The target variable, class, indicates whether a customer is high risk (bad) or low risk (good).