title	Loan Scoring Model
emoji	💰
colorFrom	red
colorTo	yellow
sdk	docker
sdk_version	29.1.4
app_file	src/api/main.py
pinned	false

Loan Scoring Model

This project presents an API that uses a machine learning model to score loan applications, and a Web Application that allows users to interact with the API and monitor the model's performance and drift.

Installation

You will need to get few services in order to use this project. They are free but needs some configuration.

Logging API request

Create an account on BetterStack:

• Create a project and a source
• Retrieve the token and the host
• Store them in .env as LOGTAIL_TOKEN and LOGTAIL_HOST (with "https//") and in Github Secrets
• Now create an API Token for this source
• And store the username, the password and the host as BETTERSTACK_USERNAME, BETTERSTACK_PASSWORD and BETTERSTACK_HOST in .env

The database should look like this:

Deploying Locally

This project uses Docker to build and run the API and Web Application. To install Docker, follow the instructions on the Docker website.

Then at the root of the project, run: docker compose up -d

Then .env should contain the following variables:

• LOGTAIL_TOKEN,
• LOGTAIL_HOST,
• BETTERSTACK_USERNAME,
• BETTERSTACK_PASSWORD,
• BETTERSTACK_HOST,
• API_BASE_URL=http://api:8000

You can then access the API at http://localhost:8000, and the Web Application at http://localhost:8050.

Web Application

The Web Application allows you two things:

• In the Demo tab, you can query the API to retrieve the information of a specific client (based on its SK_ID_CURR) and see what score the model predicts for this client,

• In the Analytics tab, you can see :
  • An Evidently Drift Report,
  • The API usage over the past day,
  • The score distribution predicted by the API over the past day,
  • The live API latency (refreshed every 5 seconds).

API

The API is a FastAPI application that uses a machine learning model to score loan applications. The API is deployed on Docker and the doc can be accessed at http://api:8000/doc.