Machine Learning with Microservices

Project Overview

This project operationalizes a Machine Learning Microservice API. It uses a pre-trained sklearn model to predict housing prices in Boston based on various features, such as the average number of rooms in a house, highway access, teacher-to-pupil ratios, and more. The data was initially sourced from Kaggle and can be found on the data source page.

The microservice serves housing price predictions via API calls. This framework can be extended to other pre-trained machine learning models, including those for image recognition or data labeling. The entire application has been containerized using Docker and deployed with Kubernetes.

File Structure

Application:

• app.py: The main application script
• model_data/: Folder containing the trained model and prediction data
• requirements.txt: Lists the dependencies for the app

Outputs:

• output_txt_files/: Folder containing output files for Docker and Kubernetes

Docker:

• Dockerfile: Defines the image for deploying the app within a container

Utils/Tools:

• Makefile: Contains useful commands for setup, installation, testing, linting, running Docker, running Kubernetes, uploading to DockerHub, and more
• run_docker.sh: Script to build and start the container
• run_kubernetes.sh: Script to deploy the app on Kubernetes
• upload_docker.sh: Script to upload the Docker image to DockerHub
• make_prediction.sh: Script to test the application and generate predictions

Setting Up the Environment

1. Create a virtual environment and activate it.
2. Run make install to install the necessary dependencies.

Running app.py

You can run the app in three different ways: Standalone, Docker, or Kubernetes.

1. Standalone:

Run the app with the following command:

python app.py

The app will be accessible at http://localhost:80.

2. Run in Docker:

Use the run_docker.sh script to build and start the container:

./utils/run_docker.sh

The script will:

• Build a Docker image
• List images to verify that the app is dockerized
• Run the container with port mapping (host port 5000 to container port 80)

The app will be accessible at http://localhost:5000.

3. Run in Kubernetes:

Use the run_kubernetes.sh script to deploy the app on a Kubernetes cluster:

./utils/run_kubernetes.sh

The script will:

• Start a container in the Kubernetes cluster (ensure you have a cluster ready, e.g., via minikube)
• Wait for the pod to start
• List the pod to confirm it is running
• Forward port 5000 (host) to port 80 (container)

The app will be accessible at http://localhost:5000.

Testing the App

To test the application, run the make_prediction.sh script. The model will return a predicted price based on the input provided. A sample input JSON is as follows:

{  
   "CHAS":{  
      "0":0
   },
   "RM":{  
      "0":9.575
   },
   "TAX":{  
      "0":296.0
   },
   "PTRATIO":{  
      "0":15.3
   },
   "B":{  
      "0":396.9
   },
   "LSTAT":{  
      "0":4.98
   }
}