Recommendation Systems

Types of Recommendation Systems

• Content‑based filtering: recommend items similar to those the user liked based on item features.
• Collaborative filtering: recommend items liked by similar users (user‑item interactions only).
• Hybrid systems: combine content‑based and collaborative signals.

Content‑Based Filtering

Represent items and user profiles in the same feature space (e.g. TF‑IDF for text) and recommend items with high similarity to the user profile.

Collaborative Filtering

• User‑based CF: find users similar to the active user and recommend items they liked.
• Item‑based CF: find items similar to items the user liked and recommend them.
• Matrix factorization: learn latent factors for users and items (e.g. SVD, ALS).

Evaluating Recommenders

Typical metrics are defined over the top‑K recommended items:

• Precision@K: proportion of recommended items in the top‑K that are relevant.
• Recall@K: proportion of all relevant items that are present in the top‑K recommendations.
• NDCG: rewards ranking relevant items higher in the list.