Can I use custom transformation functions with this skill?

Yes, the FeatureTransformer supports custom transformation functions alongside standard methods like StandardScaler, OneHotEncoding, and Box-Cox transformations.

Does it support automated documentation?

Absolutely. It integrates with SpecWeave's 'Living Docs' system to automatically generate feature definitions, transformation pipelines, and validation reports in your project folder.

Is it compatible with standard Python ML libraries?

Yes, it is designed to work seamlessly with pandas and scikit-learn ecosystems, producing model-ready dataframes and serializable transformer objects.

What types of features can be created automatically?

The skill can generate temporal features from dates, aggregation features for user behavior, interaction features between variables, and even basic NLP features from text columns.

How does this skill help prevent data leakage?

The skill includes a built-in FeatureValidator that specifically checks for perfect correlations and the presence of future data in training sets, ensuring your model results are realistic.

ML Feature Engineer

Name: ML Feature Engineer
Author: anton-abyzov

byanton-abyzov

•

データサイエンスとML

Automates the end-to-end feature engineering process for machine learning pipelines, from data quality assessment to production-ready transformations.

概要

The ML Feature Engineer skill provides a systematic framework for transforming raw datasets into high-performance features. It streamlines the entire machine learning lifecycle by automating data quality reporting, feature creation (temporal, behavioral, and interaction), dimensionality reduction via advanced selection methods, and robust transformations like scaling and encoding. Built to integrate with the SpecWeave development workflow, it ensures that every engineering step is documented, validated for data leakage or drift, and versioned for production reliability.

主な機能

Comprehensive validation for data leakage and distribution drift
Production-ready scaling, encoding, and power transformations
13 GitHub stars
Automated data quality assessment and outlier detection
Multi-method feature selection using statistical tests and model importance
Domain-specific feature creation including temporal and aggregation logic

ユースケース

Building predictive models from raw transactional or behavioral data logs
Standardizing data preprocessing for production-grade AI applications
Optimizing ML pipeline performance by reducing feature redundancy and noise

概要

主な機能

Comprehensive validation for data leakage and distribution drift
Production-ready scaling, encoding, and power transformations
13 GitHub stars
Automated data quality assessment and outlier detection
Multi-method feature selection using statistical tests and model importance
Domain-specific feature creation including temporal and aggregation logic

ユースケース

Building predictive models from raw transactional or behavioral data logs
Standardizing data preprocessing for production-grade AI applications
Optimizing ML pipeline performance by reducing feature redundancy and noise