Use Multiple Recommendation/Boosting Stages

Fusion provides three query-pipeline stages for content-based filtering and three query-pipeline stages for collaborative filtering. All of these stages recommend documents.

Content-based filtering:

Collaborative filtering:

Examples of Using Multiple Collaborative Recommendation Stages

In some cases, using multiple recommendation stages in the query pipeline might make sense. The <collection>_items_for_user_recommendations and <collection>_items_for_item_recommendations query pipelines are examples. In addition to a Recommend Items for User or Recommend Items for Item stages, each query pipeline also has a Boost with Signals stage. Thus, recommendations based on a Spark collaborative filtering computation are combined with recommendations based on the association of the query with items.

A Hybrid Recommender System

You can combine content-based filtering and collaborative filtering in the same query pipeline. A hybrid recommender system can leverage the capabilities and strengths of both approaches.