Tabular data is the bread and butter of data science projects. This repo explores XGBoost for regression which is an algorithm that is well-known by the community and is known to perform really well on tabular data. I also explored further by transforming the categorical variables using the method of Entity Embeddings which was published in a paper in 2016. Check it out on arXiv.
Private score and puplic score were retreived after submitting the predictions to Kaggle.
| Experiment ID | Categorical Variables | NaN-cats | NaN-cont | Target Transformation | Hyperparameter Search | Backtesting | Private Score | Public Score |
|---|---|---|---|---|---|---|---|---|
| 001 | Target encoder | XGBoost | XGBoost | Log transform | Default | No | 0.16925 | 0.17975 |
| 002 | Target encoder | XGBoost | XGBoost | Log transform | HyperOpt (100) | TimeSeriesSplit k = 3 | 0.13975 | 0.12481 |
| 003 | Entity Embeddings | #NAN# | FastAI | Log transform | Default | No | 0.15251 | 0.14079 |
| 004 | Entity Embeddings | #NAN# | FastAI | Log transform | HyperOpt (100) | TimeSeriesSplit k = 3 | 0.13081 | 0.11572 |
├── README.md <- The top-level README for developers using this project.
├── data
│ ├── interim <- Intermediate data that has been transformed.
│ ├── processed <- The final, canonical data sets for modeling.
│ └── raw <- The original, immutable data dump.
│
├── docs <- A default Sphinx project; see sphinx-doc.org for details
│
├── models <- Trained and serialized models, model predictions, or model summaries
│
├── notebooks <- Jupyter notebooks. Naming convention is a number (for ordering),
│ the creator's initials, and a short `-` delimited description, e.g.
│ `1.0-jqp-initial-data-exploration`.
│
├── setup.py <- makes project pip installable (pip install -e .) so src can be imported
├── src <- Source code for use in this project.
│ ├── __init__.py <- Makes src a Python module
│ │
│ ├── data <- Scripts to download or generate data
│ │ └── make_dataset.py
│ │
│ ├── features <- Scripts to turn raw data into features for modeling
│ │ └── build_features.py
│ │
│ ├── models <- Scripts to train models and then use trained models to make
│ │ │ predictions
│ │ ├── predict_model.py
│ │ └── train_model.py
│ │
│ └── visualization <- Scripts to create exploratory and results oriented visualizations
│ └── visualize.py
Project based on the cookiecutter data science project template. #cookiecutterdatascience