krishnautoml 1.0.2

Beginner-friendly AutoML library for tabular data

KrishnAutoML

KrishnAutoML is a lightweight, beginner-friendly, and production-ready AutoML library for tabular data.
It automates the end-to-end machine learning workflow with minimal user input, while keeping things modular and extensible.

---

Features

• Load data from CSV or Pandas DataFrame
• Automatic problem type detection (classification or regression)
• Smart preprocessing (missing values, categorical encoding, scaling)
• Optional EDA reports for insights
• Train multiple models (LightGBM, XGBoost, CatBoost, Scikit-Learn)
• Automated model selection and hyperparameter tuning (Optuna / GridSearchCV)
• Flexible cross-validation (KFold, StratifiedKFold, GroupKFold)
• Multiple evaluation metrics dynamically
• Early stopping and GPU support
• Save models + reproducible pipeline code
• Auto-generated reports in HTML/Markdown

---

Installation

From PyPI (after publishing):

pip install krishnautoml

From source:

git clone https://github.com/knight22-21/KrishnAutoML.git
cd KrishnAutoML
pip install -e .[dev]

---

Quick Start

Python API

from krishnautoml import KrishnAutoML

# Initialize AutoML
automl = KrishnAutoML(target="Survived", problem_type="auto")

# Full pipeline
(
    automl
    .load_data("data/titanic.csv")
    .preprocess()
    .train_models()
    .evaluate()
    .save_model("best_model.pkl")
)

print("Best model metrics:", automl.best_score)

Command Line Interface (CLI)

krishnautoml fit --data data/titanic.csv --target Survived --report

This will:

• Train models
• Save best_model.pkl
• Generate an HTML performance report

---

Example Output

Metrics (Classification example):

{'accuracy': 0.8567, 'precision': 0.8421, 'recall': 0.8312, 'f1': 0.8350}

Generated Report:

• Confusion matrix
• Feature importance
• ROC-AUC curve
• Summary of preprocessing steps

---

Advanced Usage

• Custom cross-validation:

automl = KrishnAutoML(target="SalePrice", cv_strategy="KFold", n_splits=10)

• Specify metrics:

automl = KrishnAutoML(target="Survived", metrics=["accuracy", "f1"])

• Load trained model:

from joblib import load
model = load("best_model.pkl")

---

Development

Clone and install dev dependencies:

git clone https://github.com/knight22-21/KrishnAutoML.git
cd KrishnAutoML
pip install -e .[dev]

Run tests:

pytest

Lint & format:

flake8 krishnautoml
black krishnautoml

---

License

MIT License © 2025 \Krishna Tyagi

---

Contributing

Contributions are welcome!

• Fork the repo
• Create a feature branch
• Submit a PR

---

Acknowledgements

• scikit-learn
• XGBoost
• LightGBM
• CatBoost
• Optuna