Can I use this for non-numeric data like text or images?

Yes, it covers basic text vectorization (TF-IDF, embeddings) and statistical image feature extraction techniques like texture and shape descriptors.

What is the primary benefit of the Feature Engineering skill?

It provides a structured methodology to transform raw data into high-quality features, which is often more impactful for model performance than the choice of algorithm itself.

Does this skill help prevent target leakage?

Yes, it specifically identifies target leakage as a major anti-pattern and provides validation steps to ensure features do not contain information unavailable at prediction time.

Is this skill useful for time-series analysis?

Absolutely. It includes specific patterns for extracting lag features, rolling statistics, seasonal indicators, and cyclical encoding for periodic data like hours or months.

How does it handle high-cardinality categorical data?

The skill recommends techniques beyond simple one-hot encoding, such as target encoding, frequency encoding, or embeddings, to prevent dimensionality explosion.

Feature Engineering Patterns

Name: Feature Engineering Patterns
Author: lev-os

bylev-os

0•

データサイエンスとML

Transforms raw data into optimized features to improve machine learning model performance and predictive accuracy.

This skill provides a systematic framework for the feature engineering process, enabling developers to convert raw, noisy data into high-signal variables that maximize machine learning outcomes. It guides users through exploratory data analysis, missing value strategies, categorical encoding, and domain-specific feature creation across various data types including text, geospatial, and time-series. Ideal for data scientists and ML engineers, this skill ensures best practices are followed to avoid common pitfalls like target leakage and overfitting while enhancing the representation of underlying problem structures.

主な機能

01High-cardinality encoding techniques including target and frequency encoding

02Comprehensive strategies for categorical, numeric, text, and datetime transformations

030 GitHub stars

04Feature selection and validation workflows to prevent overfitting and target leakage

05Domain-specific feature generation for finance, healthcare, and e-commerce

06Advanced handling of missing values using MCAR, MAR, and MNAR mechanisms

ユースケース

01Optimizing predictive models for industries like fraud detection and credit scoring

02Enhancing model performance when raw data doesn't align with algorithm requirements

03Creating temporal and lag features for complex time-series forecasting

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lev-os/agents feature-engineering

For use in Claude.ai and ChatGPT

Download Skill