Can I use this for custom feature engineering?

Yes, it provides templates for building custom transformers that inherit from scikit-learn's base classes for seamless integration into pipelines.

How does this skill help prevent data leakage?

It enforces the use of the Pipeline API, ensuring that preprocessing steps are learned only on training data and applied consistently during inference.

Does it support model deployment?

Absolutely. It includes patterns for model persistence using Joblib and exporting to high-speed inference formats like ONNX for production environments.

How does it improve model validation?

It provides patterns for multi-metric cross-validation and calibration curves to ensure model probabilities are reliable and training/test gaps are monitored.

Advanced scikit-learn Architecture

Name: Advanced scikit-learn Architecture
Author: tondevrel

bytondevrel

•

Data Science & ML

Implements robust scikit-learn machine learning pipelines and custom estimators to ensure production-grade, leak-free model development.

The sklearn-advanced skill empowers developers to transition from simple scripts to professional-grade machine learning architectures using the scikit-learn Pipeline API. It provides comprehensive guidance on creating custom transformers, handling heterogeneous data with ColumnTransformers, and implementing rigorous validation techniques like nested cross-validation and calibration. By enforcing the principle that 'Everything is an Object,' it eliminates common pitfalls such as data leakage and inconsistent preprocessing between training and inference, ultimately facilitating the export of models to high-performance production environments using Joblib or ONNX.

Key Features

01Custom estimator development with BaseEstimator and TransformerMixin.

02Advanced feature engineering including Target Encoding and ColumnTransformers.

03Rigorous model validation using nested cross-validation and calibration curves.

041 GitHub stars

05Production deployment patterns with Joblib and ONNX export.

06End-to-end workflow automation using the scikit-learn Pipeline API.

Use Cases

01Building complex, multi-stage feature engineering pipelines for production data.

02Performing leak-free hyperparameter tuning for high-stakes predictive models.

03Developing domain-specific custom preprocessing steps for unique datasets.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tondevrel/scientific-agent-skills sklearn-advanced

For use in Claude.ai and ChatGPT

Download Skill