MLOps-Fraud-Detection-using-Pycaret_MLFlow

PyCaret is a low-code machine learning library that allows you to quickly build and train models, while MLflow provides a robust platform for managing the entire machine learning lifecycle. Together, these tools make it easy to develop and deploy fraud detection models with confidence. By using PyCaret's built-in functionalities for data preprocessing, feature engineering, model training, and hyperparameter tuning, you can quickly build high-performing models that are optimized for fraud detection. And with MLflow's tracking and versioning capabilities, you can easily manage your experiments, reproduce your results, and collaborate with your team.

Key Tools or Libraries

1. PyCaret: a low-code machine learning library for building and training models.
2. MLflow: a platform for managing the machine learning lifecycle, including experiment tracking and versioning.
3. FastAPI: a modern, fast (high-performance) web framework for building APIs with Python.
4. Docker: a containerization platform for packaging software into portable, self-contained environments.
5. Uvicorn: a lightning-fast ASGI server for Python, used to run the FastAPI app.

API and Models

1. Model_Training.ipynb contains trainings with different processing steps like fixing imbalance, removing outliers, and feature engineering. It also contains the Experiment Tracking with MLflow.

2. Train_and_Evaluate_the_Model.ipynb contains the training of the selected model and evaluation of the best classifier.

3. Optimize_the_Model contains the code for hyper parameter tuning and voting classifier then created REST API for the model using FastAPI and create a Docker image for the API.

4. Testing_of_API This code file uses a trained model for fraud voting detection and deploys it using a Docker container. It loads data from a file, checks the Docker images and starts the API container. It then sends a request to the API endpoint for the prediction of the target variable for a particular record and prints the response. The response contains the predicted target value for the specified record.

model_utils.py The code defines a function called get_raw_data() that reads a CSV file containing credit card data, performs undersampling, and splits the data into training and testing sets. It then returns the resulting training and testing dataframes.

fraud_voting_model_api.py This script defines a FastAPI app that loads a pre-trained fraud detection model, and provides an endpoint (POST request) to make predictions on new data. The endpoints accept 30 input features related to a financial transaction and return the model's binary classification prediction of fraud or not fraud. The app is run locally on port 8001 using uvicorn. The weight file of fraud_voting_model_api.py is Weight

fraud_voting_model_api_with_docker This Python script defines a FastAPI for deploying a fraud voting model using the PyCaret classification library. It loads a pre-trained pipeline model and defines two endpoints for making predictions: POST /predict and GET /get_predict. The app runs on a Uvicorn server listening on port 8007.

Dockerfile_8007 This is a Dockerfile that defines a Python environment with specific dependencies, installs pycaret and additional packages specified in requirements2.txt, and exposes port 8007. The final command runs a Python script for deploying a fraud voting model API.

Note:

1. The port should be different from the other already existing ports. As you can see the port in Docker file is different from Dockerfile_8007 and both have different requirements.txt files.

2. The fraud_voting_model_api_with_docker.py file is different to fraud_voting_model_api.py file regarding the API endpoints, ports and host IP.