predict-aqi

Django + React application to forecast Air Quality Index (AQI) using a Long Short-Term Memory (LSTM) neural network

Overview

• Cleans historic air quality data and fill gaps in the time series using numpy and pandas
• Transforms the historic air quality time series to stationary to remove long term trend
• Creates a LSTM model for air quality forecasting using TensorFlow and scikit-learn
• Iteratively examines effect of model hyperparameters including learning rate and regularization
  • Note: The most recent training for the model yielded a RMSE of ~0.09 for the training data, and a RMSE of ~0.10 for the test data
• Utilizes Django REST Framework to create a REST API:
  • Django backend fetches current air quality data from an external API for use in forecasting air quality
• Uses React.js to fetch data from backend API and create frontend forecast interface

How to use this repository

Install Django dependencies

1. cd into the root directory
2. Run pip install -r requirements.txt

Install React dependencies

1. cd into frontend/
2. Run npm install

Run Django tests

1. cd into the root directory
2. Run python manage.py test

Run Jest + React tests

1. cd into frontend/
2. Run npm run test

Attributions

• Frontend structure based on weather app (MIT License) by @duvainel
• Background image by David Mark from Pixabay