Auto Boost

Auto Boost is a small command-line helper that mirrors the original Auto_boost.ipynb Kaggle workflow. It handles missing values, categorical encoding, cross-validated training, and submission generation for gradient-boosting models (LightGBM, XGBoost, or CatBoost) without needing to run a notebook.

Installation

From PyPI (recommended)

python -m pip install --upgrade auto_boost
# or include extras:
python -m pip install "auto_boost[lightgbm]"

From source

python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install ".[lightgbm,xgboost]"  # select any boosters you need

Base dependencies are pandas, numpy, and scikit-learn. Install at least one booster extra (lightgbm, xgboost, catboost) depending on what you plan to run.

Quickstart

auto-boost \
  --train train.csv \
  --test test.csv \
  --target Transported \
  --model-type classification \
  --metric accuracy \
  --booster lgbm \
  --folds 10 \
  --random-state 10 \
  --id-col PassengerId \
  --prediction-col Transported \
  --output submission.csv

Key flags:

• --train / --test: paths to CSV files.
• --target: column in train.csv you want to predict.
• --model-type: classification, regression, or auto to infer from the target column.
• --booster: choose between lgbm, xgb, catboost.
• --metric: auto-detected if omitted (accuracy for classification, rmse for regression).
• --output: optional CSV to save predictions (includes ID column when --id-col is supplied).

Run auto-boost --help (or auto_boost --help) for the full reference. The legacy python auto_boost.py shim has been removed in favor of the installable entrypoints.

Works With Any Tabular Dataset

• Automatic task detection when --model-type auto is supplied, so you can point the CLI at a CSV without pre-labeling it as classification vs regression.
• Smarter preprocessing that imputes instead of dropping high-cardinality categorical features and scales numerics when requested.
• Built-in label encoding for non-numeric targets (including booleans and strings), ensuring LightGBM/XGBoost/CatBoost work regardless of how the classes are represented.

Development & Packaging

For local development install in editable mode:

python -m pip install --upgrade pip build twine
python -m pip install -e ".[lightgbm]"

To produce distributable artifacts (wheel + sdist):

python -m pip install build
python -m build
ls dist/

The files under dist/ can be uploaded with twine upload dist/* when publishing to PyPI. Generated folders such as dist/, *.egg-info, and __pycache__ are ignored via .gitignore.

Releasing to TestPyPI / PyPI

Following the official Packaging Python Projects guide:

# Build fresh artifacts
rm -rf dist/
python -m build

# Upload to TestPyPI first
python -m twine upload --repository testpypi dist/*

# Verify install from TestPyPI (optional)
python -m pip install --index-url https://test.pypi.org/simple/ \
  --extra-index-url https://pypi.org/simple auto_boost

# When satisfied, push to PyPI for public install via:
# python -m pip install auto_boost
python -m twine upload dist/*

Bump auto_boost.__version__ before every upload to avoid version conflicts.

About the Notebook

The original Auto_boost.ipynb is retained for reference, but the script fixes several issues (missing class instantiation, incorrect feature-importance labels, buggy categorical handling) and is the recommended entry point for automation.