How muhammederbay10/end-to-end-loan-eligibility-prediction-system- works
Project type and purpose
• This is an end-to-end, monolithic ML web application built with Flask.
• Purpose: ingest raw loan data, preprocess it, train and serialize a model pipeline, and serve real-time loan-eligibility predictions via a web interface.
File structure
• Top-level components:
• app.py – Flask server & controller
• custom_transformers.py – preprocessing helpers
• Data/ – raw CSV datasets (train/test)
• model/ – serialized pipeline (model.joblib)
• static/ – CSS styling
• templates/ – HTML view
• requirements.txt, Procfile, runtime.txt – deployment configs
• Bank Loan approval.ipynb – exploratory & training notebook
• Patterns:
• MVC-style separation:
– Model: scikit-learn pipeline (in model/) + custom transformers
– View: HTML/CSS in templates/ + static/
– Controller: Flask routes in app.py
• Pipeline persists artifacts (joblib) and reloads at runtime.
Key insights from README
• Workflow:
1. Load CSV → clean/impute → encode/scale → train & tune multiple classifiers → select best → joblib serialize → Flask deploy
• Dependencies: pandas, scikit-learn, Flask, joblib, XGBoost
• Deployment target: Railway/Heroku via Procfile & runtime.txt