AeroHue

Demo Presentation | GitHub Repository |
Omdena Project Site

AeroHue aims to inform citizens of the AQI daily to prepare themselves better and take safety precautions in case of high AQI in Mumbai City, India.

Actively contributed in the development of the project, Monitoring and Predicting Air Quality using Machine Learning, hosted by Omdena Mumbai, India Chapter. The project has been released to the public under the name AeroHue.

My significant contributions to the project involved domain research, web scraping for data collection, exploratory data analysis (EDA), and data modeling.

The project aimed to address the impact of air pollution on human health, which is the world’s most considerable environmental risk, causing 7 million premature deaths annually. India’s air pollution is particularly severe, making it the deadliest. The community set out to monitor and predict the air quality of Mumbai City, the largest metropolitan city in India, by leveraging machine learning to explore pollution levels at a granular scale.

Data collection involved collecting data from multiple sources. Notably, past data of AQ from Kaggle, datasets from the Central Pollution Control Board, Government of India, and public APIs like WAQI, OpenAQ, data.gov.io, and OpenWeatherMap. Apart from air quality data points, weather data for the Mumbai region was also utilized.

A deep learning model was developed using the Darts library of Python to predict AQI using a given set of features such as PM2.5, PM10, SO2, CO, NOx, etc. The model predicts the AQI of Mumbai with the mean absolute percentage error (MAPE) of 3.93%

The model has been deployed on the cloud using Amazon Web Services (AWS) with a web application built on Streamlit.