This repository employs several different machine learning methods (LSTM, gradient boosting, Gaussian process regression and classification) to predict the future indoor thermal comfort performance of a naturally-ventilated test building under different future climate weather scenarios.
To-Do:
- Need to fix click arguments to ensure for any model, any list of parameters and manipulations can be passed (maybe it's better to provide the option to pass a .txt file with these specifications?)
- Need to fix 'H' returning error for CatBoost
├── LICENSE
├── Makefile <- Makefile commands
├── README.md
├── data
│ ├── interim <- Data from that has been transformed, joined, etc.
│ ├── output <- Predictions from each model.
│ ├── processed <- The final test/train data sets for used for each model.
│ └── raw <- Datafiles from variety of sources.
├── figures <- Final figures saved by user.
├── models <- Trained models.
├── src <- Directory for processing and modeling code.
│ ├── data <- Contains preprocessing functions.
│ ├── evaluation <- For producing output plots and evaluating models.
│ ├── modeling <- Specific files for data processing and training each model.
│ ├── visualization <- Exploratory visualizations of the data.
│
├── notebooks <- Jupyter notebooks for interacting with most of 'src'. Naming
│ convention is a number (for ordering), the creator's initials,
│ and a short `-` delimited description, e.g.
│ `1.0-jqp-initial-data-exploration`.
│
├── references <- Data dictionaries, manuals, and all other explanatory materials.
├── requirements.txt <- The requirements file for reproducing the analysis environment, e.g.
generated with `pip freeze > requirements.txt`