This repository contains the Jupyter notebooks and deployment infrastructure/scripts used in the COMP3000 project to predict future air quality and temperature.
%%{ init: { 'flowchart': { 'curve': 'stepAfter' } } }%%
graph
aggregator[Aggregator API]
frontend[Frontend application]
shim[DEFRA Shim service]
csvs[DEFRA CSVs]
predictions[Prediction service]
metadata[AURN station metadata]
aggregator --- |Historical temperature data| shim
aggregator --- |Historical PM2.5 data| csvs
aggregator --- |Future predictions from trained model| predictions
aggregator --- |Station metadata| metadata
frontend --- aggregator
- Frontend
- Aggregator
- Predictions (you are here)
- Metadata
- Shim
This service uses Jupyter Notebooks and Tensorflow to produce machine learning models based on pyaurn data. These models are hosted in OneDrive, where they are pulled and unpacked in the Dockerfile which builds on Tensorflow Serving.
Training can be performed by installing Tensorflow and running the notebooks as any other.
Hosting can be performed by building the dockerfile, which will pull the current models from OneDrive.
- For: researchers and the public
- Who: need a better understanding of the climate crisis
- The: environmental data dashboard
- Is a: web application dashboard
- That: can make predictions for the future of the environment and display environmental data such as air pollution in an easy-to-understand format