Azure Machine Learning Project: Air Quality Classification

Project Overview

This project addresses a classification problem related to air quality. It demonstrates the complete end-to-end process of building, optimizing, deploying, and monitoring a machine learning model using Azure Machine Learning. The project involves creating a robust machine learning pipeline, utilizing tools such as Optuna for hyperparameter tuning, MLflow for experiment tracking, and deploying the model as a REST API with an optimum front-end interface for real-time predictions.

---

Key Features

1. Data Exploration and Preprocessing

   • Dataset sourced from public platforms like Kaggle or UCI.
   • Data preprocessing: handling missing values, encoding categorical variables, scaling numerical features.

2. Model Training and Optimization

   • Built initial model using algorithms such as Random Forest and XGBoost.
   • Hyperparameter tuning with Optuna or Azure ML HyperDrive.
   • Experiment tracking using MLflow.

3. Model Deployment

   • Deployed as a REST API endpoint on Azure Machine Learning.
   • Developed a scoring script (score.py) for endpoint inference logic.

4. Front-End Interface

   • Created an interactive application using Streamlit (app.py) for user input and real-time predictions.

5. Monitoring and Evaluation

   • Monitored model performance, tracked latency, and logged metrics using Azure tools.

---

Repository Contents

• data_exploration_to_deployment.ipynb: Step-by-step Jupyter Notebook for data exploration, preprocessing, model training, and deployment.
• data_exploration_to_deployment.py: End-to-end pipeline script for setting up the environment, training the model, and deploying it.
• pipeline_logs.txt: Execution logs from the Azure compute instance.
• score.py: Scoring script for handling model inference requests at the deployed endpoint.
• app.py: Streamlit app for real-time user interaction and prediction visualization.
• test_input.json: Sample input JSON for testing the deployed endpoint.
• requirements.txt: Installation of the Python dependencies (libraries) used.
• Azure Machine Learning Air Quality Classification Project.pdf: Detailed project report documenting all phases, challenges, and results.

---

Getting Started

Prerequisites

• Azure subscription with Azure Machine Learning resources configured.
• Python 3.8 or later with the required libraries installed (requirements.txt provided in the project).

Setup and Execution

1. Clone the repository:

   git clone <repository_url>
   cd <repository_folder>

2. Install dependencies:

   pip install -r requirements.txt

3. Run the pipeline script:

   python data_exploration_to_deployment.py

4. Start the Streamlit app:

   streamlit run app.py

5. Test the deployed model endpoint using test_input.json.

---

This project was developed from scratch under time constraints. As a result, it is not perfect and may contain errors or potential improvements. If you find it useful, feel free to create an issue to share your feedback with me.

Thank you!