underdoge/fraud-detection

Final Project for the Código Facilito's 2023 Machine Learning Bootcamp

Fraud Detection 🔎

Final Project for the Código Facilito's 2023 Machine Learning Bootcamp

Hosted Streamlit App

A hosted version of the model can be found on Streamlit here.

Installation

Open up a Terminal and enter the following commands:

Cloning the repository

git clone https://github.com/underdoge/fraud-detection

cd fraud-detection

Creating the virtual environment

python -m venv .venv

Activating the virtual environment

source .venv/bin/activate

Installing on macOS / Linux

pip install -r requirements.txt

Launching the mlflow server

mlflow server

Running the metaflow pipeline (requires mlflow server running)

python fraud_detection_flow.py run --source-file data/credit_card_transactions-ibm_v2.csv

Import mlflow model into bentoml

python import_mlflow_model.py

Verify model was successfully imported in bentoml

bentoml models list

Launch bentoml service

cd service
bentoml serve service.py --reload

Build Bento

bentoml build

List created Bentos

bentoml list

Create Docker image from Bento (requires Docker)

bentoml containerize <model tag from previous step e.g. fraud-detection-service:worn7ggjg2q63yqs>

Run model from Docker image (requires Docker)

docker run -p 3000:3000 <model tag from previous step e.g. fraud-detection-service:worn7ggjg2q63yqs>

Requirements

• Python 3.11.6 or greater
• Git (to clone the repo)
• Docker (to create Docker image)

Dataset Sources

• Credit Card Transactions Kaggle Dataset