Deploy a simple supervised machine learning algorithm in Azure as a function.

This post outlines the development of a supervised machine learning model using the DecisionTreeClassifier algorithm. We will cover training the model on a specific dataset, fine-tuning its parameters, and evaluating accuracy. After achieving proficiency, we’ll deploy the model as an Azure Function for serverless execution and expose it as a REST API for real-time predictions. This approach offers insights into machine learning and cloud-based solutions.

Note:Azure provides machine learning services through Azure ML Designer, which is a visual drag-and-drop interface for building and deploying machine learning pipelines, and Azure ML SDK, which allows for programmatic control and customization.

🚀 Features

• Machine Learning Model: DecisionTreeClassifier trained on the famous Iris dataset
• REST API: FastAPI-powered endpoints with automatic documentation
• Cloud Deployment: Serverless deployment on Azure Functions
• Containerization: Docker support for consistent environments

🏃‍♂️ Quick Start

Prerequisites

• Python 3.8+
• pip or conda
• Docker (optional)
• Azure CLI (for Azure deployment)

Local Installation

1. Clone the repository

   git clone https://github.com/uday160386/ml-model-as-rest-api
   cd ml-model-deployment
   

2. Install dependencies

   pip install -r requirements.txt
   

3. Run the application

   uvicorn main:app --reload
   

4. Access the application
   • API: http://localhost:8000
   • Interactive docs: http://localhost:8000/docs
   • Alternative docs: http://localhost:8000/redoc

     🌺 Dataset Information

Iris Dataset from scikit-learn:

• Samples: 150
• Features: 4 (sepal length, sepal width, petal length, petal width)
• Classes: 3 (Setosa, Versicolor, Virginica)
• Type: Multiclass classification

The dataset is automatically loaded and preprocessed when training the model.

📁 Project Structure

ml-model-deployment/
├── main.py                 # FastAPI application
├── model/
│   ├── __init__.py
│   ├── train.py           # Model training script
│   └── iris_model.joblib  # Trained model file
├── requirements.txt       # Python dependencies
├── Dockerfile            # Docker configuration
└── README.md           # This file

🧪 Testing

Manual Testing

Use the provided examples in the interactive documentation or test with curl:

curl -X POST "http://localhost:8000/predict" \
-H "Content-Type: application/json" \
-d '{
  "sepal_length": 6.3,
  "sepal_width": 2.5,
  "petal_length": 5.0,
  "petal_width": 1.9
}'

📊 Model Performance

• Accuracy: ~96% on test set
• Training Time: < 1 second
• Prediction Time: < 10ms per request
• Model Size: ~2KB

📚 API Documentation

Endpoints

POST /predict

Predict the Iris species based on flower measurements.

Request Body:

{
  "sepal_length": 5.1,
  "sepal_width": 3.5,
  "petal_length": 1.4,
  "petal_width": 0.2
}

Response:

{
  "prediction": "setosa",
  "confidence": 0.95,
  "all_probabilities": {
    "setosa": 0.95,
    "versicolor": 0.03,
    "virginica": 0.02
  }
}

🐳 Docker Deployment

docker run -p 8000:8000 venmaum/ml-simple-app

☁️ Azure Functions Deployment

Prerequisites

• Azure CLI installed and configured
• Azure Functions Core Tools

Deployment Steps

1. Login to Azure

   az login
   

2. Create Resource Group

   az group create --name ml-app-rg --location eastus
   

3. Deploy Function App

   func azure functionapp publish <function-app-name>
   

4. Test the deployed function

   curl -X POST https://<function-app-name>.azurewebsites.net/api/predict \
   -H "Content-Type: application/json" \
   -d '{"sepal_length": 5.1, "sepal_width": 3.5, "petal_length": 1.4, "petal_width": 0.2}'
   

📚 References

• Scikit-learn Decision Trees
• FastAPI Documentation
• Azure Functions Python Guide
• Source Code

🏷️ Tags

machine-learning fastapi azure-functions docker iris-dataset decision-tree rest-api python scikit-learn