AirSight India – AQI Prediction System

About the Project

AirSight India is a web application that predicts the Air Quality Index (AQI) for Indian cities using machine learning.

The project automatically downloads the latest AQI bulletin from the CPCB website every day, processes the data, updates the dataset, and provides AQI predictions through a web application.

I built this project to understand how an end-to-end machine learning application works, from collecting data to deploying a prediction system.

Features

Automatic daily AQI data collection from CPCB
PDF data extraction and preprocessing
Feature engineering using historical AQI data
AQI prediction using an XGBoost model
FastAPI backend for prediction APIs
React-based frontend dashboard
Seven-day AQI forecast
Health impact indicators
Automated daily data pipeline using GitHub Actions

Tech Stack

Frontend

React
Vite
Tailwind CSS

Backend

FastAPI
Python

Machine Learning

XGBoost
Pandas
NumPy
Scikit-learn

Automation

GitHub Actions

Model Performance

Model: XGBoost Regressor

Evaluation: 80/20 Train Test Split

R² Score: 0.8387
MAE: 13.64 AQI

What I Learned

Working on this project helped me gain practical experience with:

Building complete ML pipelines
Data preprocessing and feature engineering
Developing REST APIs using FastAPI
Frontend development using React
Automating workflows with GitHub Actions
Integrating machine learning models into web applications

Future Improvements

Automatic model retraining after daily data updates
Weather data integration
Improved long-term forecasting
More visualization and analytics features