How does this skill handle the 'cold start' problem?

It provides implementation patterns for fallback chains, including popularity-based recommendations, demographic-based suggestions, and exploration strategies like epsilon-greedy.

Does it support evaluation metrics for ranking quality?

Absolutely. It includes implementation logic for NDCG (Normalized Discounted Cumulative Gain), MAP@K, and Coverage to ensure your recommendations are high-quality and diverse.

What recommendation algorithms are included?

This skill supports user-based and item-based collaborative filtering, matrix factorization via SVD, and hybrid models that combine different approaches for better accuracy.

Can I use this for very sparse datasets?

Yes, it specifically includes matrix factorization techniques like SVD which are designed to handle matrices that are over 99% empty by discovering latent factors.

Recommendation Engine Builder

Name: Recommendation Engine Builder
Author: secondsky

bysecondsky

•

データサイエンスとML

Implements advanced recommendation systems using collaborative filtering, matrix factorization, and hybrid algorithms for personalized user experiences.

概要

The recommendation-engine skill enables Claude to architect and build sophisticated personalization systems for apps and platforms. It covers a wide spectrum of approaches, including user-based and item-based collaborative filtering, matrix factorization (SVD) for sparse data, and hybrid models that combine content-based features with interaction history. Beyond simple modeling, this skill provides production-grade solutions for common industry challenges like the cold-start problem, popularity bias, and data sparsity, while offering robust evaluation frameworks using metrics like NDCG, Precision@K, and MAP@K to ensure high-quality suggestions.

主な機能

Collaborative Filtering (User-based and Item-based)
Hybrid Recommender logic for combining content and collaborative signals
Matrix Factorization (SVD and ALS) for handling sparse datasets
Advanced ranking evaluation metrics like NDCG and Precision@K
21 GitHub stars
Cold Start strategies for new users and items

ユースケース

Developing 'People Also Viewed' cross-selling features for SaaS or retail
Creating content discovery engines for media and news platforms
Building personalized 'Recommended for You' product carousels for e-commerce

概要

主な機能

Collaborative Filtering (User-based and Item-based)
Hybrid Recommender logic for combining content and collaborative signals
Matrix Factorization (SVD and ALS) for handling sparse datasets
Advanced ranking evaluation metrics like NDCG and Precision@K
21 GitHub stars
Cold Start strategies for new users and items

ユースケース

Developing 'People Also Viewed' cross-selling features for SaaS or retail
Creating content discovery engines for media and news platforms
Building personalized 'Recommended for You' product carousels for e-commerce