GDZ Elektrik

Overview

This project contains scripts and utilities developed for the GDZ Elektrik Datathon 2024, aimed at forecasting unintended power outages. Our approach integrates multiple models including XGBoost, CatBoost, and LightGBM, leveraging extensive feature engineering and data preprocessing to enhance prediction accuracy.

Competition Link

For more details about the competition, datasets, and evaluation metrics, visit the Kaggle competition page.

Repository Structure

experiments/
│   ├── <timestamp>_submission.csv          - Predictions ready for submission.
│   ├── <timestamp>_val_plot.png            - Visualization of validation predictions.
│   ├── <timestamp>_yengecburger_kitchen.py - Script containing main logic for feature engineering and model training.
│   ├── <timestamp>_yengecburger_v4.py      - Utility scripts for various preprocessing tasks.
│   └── <timestamp>_log.txt                 - Logs generated during the experiment.

Running the Experiments

1. Setup Environment: Ensure Python 3.8+ is installed and set up a virtual environment.

   python -m venv yengecburger_env
   source yengecburger_env/bin/activate  # On Windows use `yengecburger_env\Scripts\activate`

2. Install Dependencies:

   pip install pandas numpy matplotlib seaborn xgboost catboost lightgbm scikit-learn scipy

3. Execute the Script:

   Navigate to the script's directory and run:

   python yengecburger_kitchen.py

   This will execute the experiment, logging the output and results in the experiments folder.

Viewing Results and Images

Results from the experiments, including prediction accuracy and feature importance plots, are saved in the experiments directory.

Sample plot below:

Additional Notes

• Ensure that all data files from the Kaggle competition are placed in the correct directories as expected by the scripts.
• Modify the config_params in the yengecburger_kitchen.py script to tune the experiment settings.

---

Old Logs:

v20 3.0303951367781155 #num of valiadion windows 4 hyper params:default v21 3.027355623100304 #num of valiadion windows 8 hyper params:default v22 3.020516717325228 #num of valiadion windows 16 hyper params:default

v23 3.0364741641337387 #num of valiadion windows 4 hyper params:light v24 3.234042553191489 #num of valiadion windows 8 hyper params:light v25 3.0372340425531914 #num of valiadion windows 16 hyper params:light

v26 2.947568389057751 #num of valiadion windows 4 hyper params:default refit_full=TRUE v27 2.9703647416413372 #num of valiadion windows 8 hyper params:default refit_full=TRUE v28 2.981003039513678 #num of valiadion windows 16 hyper params:default refit_full=TRUE

v29 2.9506079027355625 #num of valiadion windows 2 hyper params:default refit_full=TRUE v30 3.1603343465045595 #num of valiadion windows 1 hyper params:default refit_full=TRUE