Radon level forecasting using LSTM models

This work builds a simple LSTM model that it is able to predict the radon levels using only the radon itself and the ventilation status in the room it is monitorized. It has below 15Bq/m3 of RMSE and the next figure shows a the model forecasting prediction.

The repository structure is the following

├── data
│   ├── predictions.csv
│   └── radon-data.csv
├── figures
│   └── // figures used, drawed with src/plots.R
├── README.md
└── src
    ├── LSTM-models.ipynb
    ├── plots.R
    ├── requirements.txt
    └── utils
        └── // Useful functions

Depoyment with docker

Start the container as follows:

docker build -t radon_predictions .
docker run -p 8500:8500 --name radon radon_predictions

To submit data to predict new radon levels, there is an example in sample_predictions.py using python requests.

Then, to stop the container: docker stop radon. For further configuration, navigate the docker documentation.