Can I use this for imbalanced datasets?

Absolutely. It contains specific patterns for the 'themis' package, including SMOTE, ADASYN, and downsampling techniques to handle class imbalance.

What is the primary focus of this Claude Code skill?

The skill provides standardized, high-quality code patterns for feature engineering and data preprocessing using the R 'recipes' package within the tidymodels framework.

How does this skill help prevent data leakage?

It implements the 'prep' and 'bake' workflow, ensuring that all statistical parameters (like means or variances) are calculated only from training data and then applied to test data.

Does it support advanced dimensionality reduction?

Yes, it includes patterns for PCA, UMAP (via the embed package), and Kernel PCA to help reduce feature space complexity.

Is there support for text data feature engineering?

Yes, it provides patterns using the 'textrecipes' package for tokenization, stopword removal, stemming, and TF-IDF calculation.

Tidymodels Recipe Patterns

Name: Tidymodels Recipe Patterns
Author: choxos

bychoxos

データサイエンスとML

Streamlines feature engineering in R using standardized preprocessing patterns for the recipes package.

概要

This skill provides a comprehensive library of feature engineering patterns for the R recipes package, part of the tidymodels ecosystem. It enables data scientists to build robust, reproducible preprocessing pipelines for machine learning by providing snippets for normalization, imputation, categorical encoding, and dimensionality reduction. By following these established patterns, users can ensure best practices such as preventing information leakage and maintaining proper step ordering (Impute → Transform → Encode → Normalize) to create high-quality models that generalize well to new data.

主な機能

0 GitHub stars
Comprehensive numeric preprocessing including normalization, splines, and Yeo-Johnson transformations
Advanced categorical encoding techniques like target encoding, dummy variables, and handling rare levels
Dimensionality reduction patterns for PCA, UMAP, and Kernel PCA
Multiple imputation strategies ranging from simple mean/mode to KNN and bagged trees
Specialized handling for text data, date/time features, and class imbalance (SMOTE/ADASYN)

ユースケース

Managing missing data and class imbalance in large-scale tabular datasets
Building repeatable data preprocessing pipelines for tidymodels machine learning workflows
Implementing complex feature engineering for high-cardinality categorical variables using target encoding

概要

主な機能

0 GitHub stars
Comprehensive numeric preprocessing including normalization, splines, and Yeo-Johnson transformations
Advanced categorical encoding techniques like target encoding, dummy variables, and handling rare levels
Dimensionality reduction patterns for PCA, UMAP, and Kernel PCA
Multiple imputation strategies ranging from simple mean/mode to KNN and bagged trees
Specialized handling for text data, date/time features, and class imbalance (SMOTE/ADASYN)

ユースケース

Managing missing data and class imbalance in large-scale tabular datasets
Building repeatable data preprocessing pipelines for tidymodels machine learning workflows
Implementing complex feature engineering for high-cardinality categorical variables using target encoding