Loan Eligibility Predictor
A credit decision from applicant data: XGBoost against a logistic baseline, served by Flask and shipped by CI into a container.
FBN.2025.15
Loan Eligibility Predictor
Machine learning engineer, sole build
Python, scikit-learn, XGBoost, Flask, 2025
2
models compared, XGBoost and logistic
5
evaluation measures, ROC-AUC included
CI
tested and containerised on every push
The problem
A loan decision should be fast, explainable and reproducible, not a spreadsheet and a hunch.
The approach
An end-to-end pipeline of preprocessing, feature engineering and training that pits XGBoost against a logistic-regression baseline, measured on accuracy, precision, recall, F1 and ROC-AUC, served through a Flask API and web interface, tested and containerised by a GitHub Actions pipeline.
The outcome
A live prediction service with its evaluation, tests and container image all built by CI.
What it does
3 capabilitiesA baseline to beat
XGBoost earns its place against logistic regression on five measures.
From notebook to service
A Flask API and web interface put the model in front of users.
Shipped by a pipeline
GitHub Actions tests the code and builds the Docker image.
The cabinet
Captured from the live build
Loan Eligibility Predictor, plate I
Screen capture, 2560 × 1600
ml-loan.onrender.com, MMXXVI
Next exhibit —
IEEE Tournament Platform