liveBook · Manning

Session-based recommendation

As explained in AI Recommender Systems

A session-based recommendation selects items from the sequence of interactions occurring within a user's current session.

Session-based recommendation selects items from interactions occurring during a user's current session. It is most useful when a user has some activity but limited long-term history, providing immediate behavioral signal before collaborative filtering becomes reliable. section 4.7

Use session-based recommendation to:

  • Blend current-session behavior with popularity and content-based filtering for users who have only a handful of interactions. section 4.7
  • Encourage discovery across categories instead of narrowing recommendations too early. section 4.7
  • Treat collaborative filtering cautiously until the user's history provides enough signal. section 4.7
Figure 4.19 The different stages of a user's lifecycle, and which types of recommendations can be used at each stage.

Use the lifecycle figure to place session-based recommendation alongside other methods across stages of user activity. section 4.7

How AI Recommender Systems covers it

How deep it goes
Positions the technique within a lifecycle strategy but does not specify algorithms, data structures, or production implementation.
Examples
Uses users with handfuls of interactions as the concrete lifecycle example and recommends cross-category discovery.
What it assumes
Assumes interaction tracking and distinguishes users with current activity from users with established histories.
What you can do after
Apply it by blending current activity with popularity and content signals while delaying noisy collaborative filtering.
How it connects
Connects session-based recommendation with popularity, content-based filtering, collaborative filtering, and context-based recommendations.

This explainer was generated by Manning's knowledge graph from the text of AI Recommender Systems. The section links point to the parts of the book it draws on.