Skip to main content

Rising Popularity (Rule-Based)

Description​

The Rising Popularity policy ranks or scores items based on their recent increase in popularity or engagement momentum. It identifies items that are rapidly gaining traction over a defined time period, emphasizing the rate of change. Useful for surfacing viral content or breaking news.

Policy Type: rising-popularity Supports: scoring_policy

Hyperparameter tuning​

  • event_values: List of event value strings to filter interactions by.
  • time_window: Time window for measuring popularity change.
  • time_frequency: How often the score updates (30M, 1H, 1D, or 1W).

V1 API

policy_configs:
  scoring_policy:
    policy_type: rising-popular
    # Parameters define how trend/momentum is calculated. 
    # Example parameters:
    time_window: 7       # Time window for measuring popularity change (verify syntax/unit)
    time_frequency: "1D"    # How often the score updates (verify syntax/unit)

Usage​

Use this model when:​

  • You want to surface items that are rapidly gaining engagement or momentum
  • You need a non-personalized trending or “what’s hot now” experience
  • You are dealing with cold-start users where personalized history is limited or unavailable
  • You want a simple, robust baseline for time-sensitive content without training a complex model

Choose a different model when:​

  • You need personalized recommendations that strongly depend on user history (use ALS, ELSA, Two-Tower, or sequential models)
  • You care more about long-term relevance than short-term spikes in popularity (use learned scoring models like LightGBM, XGBoost, DeepFM, or Wide & Deep)
  • Your catalog changes slowly and recent engagement is not a strong signal
  • You need fine-grained control over feature interactions and ranking objectives (use learning-to-rank or neural scoring models)

Use cases​

  • Trending or “Top rising” sections on homepages and category pages
  • Surfacing viral or breaking news articles, videos, or social posts
  • Cold-start experiences for new or anonymous users before personalization kicks in
  • Fallback or blended signals in multi-signal ranking strategies (e.g., mix of trending and personalized)

Reference​

  • Algorithm depends on implementation (e.g., time-series analysis, windowed comparisons). No single standard paper.