Urban temperarure correction model (B&C)

This repository contains code used to conduct modelling of data collected as part of a study on building facade temperatures in relation to meteorological data.

Contents

BCformatter.py processes data stored in ./data/data.xslx and returns a formatted dataset in ./data/training_data.csv. BCtrain.py is responsible for training and contains code for some initial data preprocessing. To visualise model ouputs, BCgraph.py is used to render plots. BCshap generates explainer value plots.

How to run

Make sure you have python 3.+ and pip installed. Clone the repository. Navigate to the created folder and in terminal, run the following commands (linux/macos):

python -m venv .venv

then

source .venv/bin/activate

and

pip install -r requirements.txt

to run predictions/training on GPU, install tensorflow extensions; eg to run on Apple Silicon GPU, install tensorflow-metal.