DVG Climate Chamber Comfort dataset

Project Overview

This project focuses on analyzing climate chamber data, preprocessing it, and building machine learning models to predict thermal preference based on the data collected during a thermal comfort study at the climate chamber of RWTH Aachen University, Germany. The repository includes notebooks for Exploratory Data Analysis (EDA), scripts for model training, result analysis, and pre-trained models.

Repository Structure

DVG-CLIMATE-CHAMBER-COMFORT-DATASET/
│
├── data/                                        # Directory containing datasets and information about the experiment
│   ├── ...                                 
│   └── README.md                                   # Readme file for the dataset
│
├── EDA_notebooks/                               # Notebooks for data preprocessing and exploratory data analysis
│   ├── preprocessing.ipynb                         # Notebook for data cleaning and preprocessing
│   └── rawdata_analysis.ipynb                      # Exploratory analysis of raw data
│
├── EDA_plots/                                   # Directory for storing generated EDA plots and visualizations
│
├── model_scripts/                               # Scripts for model training and evaluation
│   ├── config.py                                   # Configuration file for Gridsearch cv parameters and feature sets
│   ├── model_train.py                              # Script for training machine learning models
│   └── plot_results.ipynb                          # Notebook for generating and analyzing result plots
│
├── results/                                     # Saved models and output results
│   ├── ET_best_model-F_accessible.pkl              # ExtraTrees model trained on f_accessible feature sets
│   ├── ET_best_model-F_all.pkl                     # ExtraTrees model trained on f_all feature sets
│   ├── ET_best_model-F_fanger.pkl                  # ExtraTrees model trained on f_fanger feature sets
│   ├── model_results.csv                           # Model performance results in CSV format
│   ├── model_results.json                          # Model performance results in JSON format
│   ├── pred_results.json                           # Model predictions in JSON format
│   ├── SVC_best_model-F_accessible.pkl             # SVC model trained on f_accessible feature sets
│   ├── SVC_best_model-F_all.pkl                    # SVC model trained on f_all feature sets
│   └── SVC_best_model-F_fanger.pkl                 # SVC model trained on f_fanger feature sets
│
├── results_plots/                               # Plots generated from model predictions and evaluations
│
├── requirements.txt                             # Python dependencies needed to run the project
│
├── LICENSE.mos                                  # MIT License file 
│
└── README.md                                    # Project documentation

How to use

Clone the Repository

git clone https://github.com/TUB-DVG/DVG-climate-chamber-comfort-dataset.git
cd DVG-climate-chamber-comfort-dataset

Dependencies

Ensure you have Python 3.8+ installed. The project uses the following Python libraries:

• numpy
• pandas
• scikit-learn
• matplotlib
• seaborn

Install the required dependencies:

pip install -r requirements.txt

Data Preprocessing and Analysis

• Raw Data Analysis: Use rawdata_analysis.ipynb for initial EDA.
• Preprocessing: Run the preprocessing.ipynb notebook to clean and preprocess the raw data.

Model Training

The script model_train.py trains machine learning models (ExtraTrees and SVC). The results are saved in the results/ folder as .pkl files for pre-trained models and plot_results.ipynb to visualize model performance.

-data_path: Path to the processed data file

-output_dir: Directory where model outputs and results will be saved

Run the following script in the model_script directory

python model_scripts/model_train.py --data_path ./data/processed_data.csv --output_dir ./results

Usage

• Feature Sets: The project uses three predefined feature sets for training the models. When using the pre-trained models or running predictions, input features must match one of these formats:

  -F_fanger: ['rt', 'mrt', 'v', 'rh', 'clo', 'met']

  -F_all: ['met', 'age', 'sex', 'clo', 'CO2', 'rh', 'rt','out_t', 'out_rh', 'w_st', 'a_st', 'hr', 'v', 'bmi']

  -F_accessible: ['met', 'age', 'sex', 'rh', 'rt', 'out_t', 'out_rh', 'w_st', 'hr', 'bmi']

The pre-trained models can be imported as follows:

import pickle
import pandas as pd

# Load a pre-trained model
model_path = "results/ET_best_model-F_accessible.pkl"  
with open(model_path, "rb") as file:
    et_model = pickle.load(file)

# Example input for f_accessible
input_data = [[1.2, 30, 'Female', 39.77, 18.63, 1.336, 94.70, 28.582, 86.94, 22.83]]

# Define column names
cols = ['met', 'age', 'sex', 'rh', 'rt', 'out_t', 'out_rh', 'w_st', 'hr', 'bmi']

# Convert to DataFrame
df = pd.DataFrame(input_data, columns=cols)

# Run prediction (0:No change; -1:Cooler; 1:Warmer)
prediction = et_model.predict(df)
print("Prediction:", prediction)

License

This project is licensed under the MIT License.

Acknowledgments

Special thanks to RWTH Aachen University for providing access to the climate chamber facilities and supporting data collection efforts. Also to the Einstein Center Digital Future (ECDF), Berlin for funding this project.

Contact

[email protected]

[email protected]