Data Science Toolkit

DSToolkit is a modular Python library designed to accelerate applied data science workflows.
It provides AutoML pipelines, model evaluation utilities, interpretability tools, and feature engineering helpers for:

• Classification
• Regression
• Clustering

The library follows a scikit-learn–inspired API, focusing on:

• Reproducibility
• Clean abstractions
• Practical diagnostics for real-world ML problems

Key Features

AutoML

• Automated training pipelines for:
  • Classification
  • Regression
  • Clustering
• Built-in support for:
  • LightGBM
  • CatBoost
  • HistGradientBoosting
  • KMeans
  • Gaussian Mixture Models
• Hyperparameter optimization
• Holdout and cross-validation strategies

Model Analysis & Metrics

• Classification metrics:
  • ROC, PR, KS, calibration curves
  • Custom scorers and lift-based metrics
• Regression diagnostics:
  • Residual analysis
  • Error by quantile
  • True vs predicted
• Clustering evaluation:
  • Silhouette analysis
  • Cluster size distribution
  • Feature statistics per cluster

Interpretability

• SHAP-based explanations
• Feature importance
• Decision tree surrogates (OvR and OvO)

Feature Engineering

• Encoders and wrappers compatible with sklearn pipelines
• Custom transformation utilities

Project Structure

src/dstoolkit/
├── automl/                # AutoML pipelines
├── feature_engine/        # Feature engineering utilities
├── metrics/               # Custom metrics and plots
├── model/
│   ├── analysis/          # Model diagnostics and visualization
│   └── interpretability/  # Explainability tools
├── preprocessing/         # Data preprocessing helpers

Example notebooks demonstrating usage can be found in the notebooks/ directory.

Instalation

pip install -r requirements.txt

Or, for development:

pip install -e .

Quick Example

from dstoolkit.automl.classifier import AutoMLLightGBMClassifier

automl = AutoMLLightGBMClassifier(
    scoring="roc_auc",
    n_trials=50
)

automl.fit(X_train, y_train)
automl.evaluate(X_test, y_test)

Documentation

Detailed documentation is available in the docs/ directory:

• AutoML APIs
• Metrics and scoring
• Model analysis and interpretability
• Feature engineering utilities

License

This project is licensed under the MIT License - see the LICENSE file for details