skills.rest
thbraet
thbraetCOMMUNITY
Explore all skills by this author →
GitHub

select-features

Select optimal engineered feature subsets using forward selection and overfit gap monitoring.

Updated Mar 20, 2026
Search Tags:#notebook#overfitting#data-preparation#feature-selection#crisp-dm#model-generalization#forward-selection
One-click install
npx skills add https://github.com/thbraet/claude-template --skill select-features
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: select-features
Source: https://github.com/thbraet/claude-template/tree/main/skills/select-features
Command: npx skills add https://github.com/thbraet/claude-template --skill select-features

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Select-features reduces dimensionality and prevents overfitting by identifying the most predictive engineered features, improving model generalization and reducing training time.

Core Features & Use Cases

  • Feature group analysis: evaluates logical feature groups (core, temporal, domain-specific) to identify signal vs. noise.
  • Forward selection with monitoring: greedily adds features while tracking CV accuracy and overfit gap to avoid overfitting.
  • Notebook and report generation: outputs a feature selection notebook and a structured summary documenting results and recommendations.

Quick Start

Run the feature-selection steps on your engineered feature set to identify the optimal subset for modeling.

Frequently Asked Questions about select-features

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I select features to prevent overfitting on a small dataset?▼

To prevent overfitting on small datasets, apply forward feature selection to iteratively add features while monitoring the cross-validation accuracy and overfit gap. This identifies the optimal subset that maintains generalization without introducing noise.

What is forward selection and how does it help with model generalization?▼

Forward selection is a greedy feature selection method that iteratively adds features to a model. It improves model generalization by evaluating feature groups and stopping when additional features increase the overfit gap.

How do I evaluate feature groups to separate signal from noise?▼

Evaluate feature groups by analyzing logical categories like core, temporal, and domain-specific features. This group analysis identifies true predictive signal versus noise, reducing dimensionality and improving model generalization.

When should I use forward selection instead of other feature selection methods?▼

Use forward selection for small to medium datasets where the feature-to-sample ratio is a concern. It is particularly effective when you need to monitor the overfit gap and evaluate logical feature groups to determine when to stop adding features.

How do I document the rationale for dropping features during selection?▼

Document dropped features by generating a structured summary alongside a feature selection notebook. This report records evaluation results, including the rationale for why specific feature subsets were selected or excluded to prevent overfitting.

Related Skills

Explore More Data & Analytics Skills→
snoodleboot-iosnoodleboot-io
2

feature-engineering-minimal

Educate on feature selection, normalization, and dimensionality reduction for machine learning models.

Official
Basic
secondskysecondsky
204

ml-model-training

Train machine learning models with scikit-learn, PyTorch, and TensorFlow.

Community
Advanced
thbraetthbraet

assess-model

Assess trained models beyond aggregate metrics for CRISP-DM workflows.

Community
Advanced
tassiovaletassiovale
13

umap-learn

Reduce high-dimensional data to lower-dimensional embeddings with umap-learn.

Community
Advanced