Online Game Friend Recommendation System Using Player Behavior Analysis

The patent segments the friend recommendation system into distinct functional modules: a complex event engine that processes player behavior data, a player model generator that creates behavioral profiles, and a recommendation engine that matches players based on segmented criteria such as playing habits, communication patterns, and game preferences. This modular segmentation allows each component to handle specific aspects of the matching process independently, improving match quality while managing system complexity through division of labor.

The patent introduces player models as intermediary representations that mediate between raw player data and the recommendation process. These player models serve as intermediate artifacts that capture essential behavioral patterns and preferences, allowing the system to match players based on processed insights rather than raw data directly. This intermediary layer improves matching reliability by providing structured, interpretable representations while managing complexity through abstraction.

The patent implements preliminary action by continuously generating and updating player models in the background as players engage with the game, so that when friend recommendations are needed, pre-processed behavioral data and player models are already available. This allows the recommendation engine to query pre-computed player models rather than processing raw event data in real-time, improving recommendation accuracy through up-to-date player profiles while reducing computational load during the actual recommendation moment.

The patent establishes continuous data collection and player model updating that runs parallel to game gameplay, ensuring that player behavioral profiles are continuously refined without interrupting the gaming experience. This continuous background processing maintains high recommendation accuracy by keeping player models current while distributing computational load over time rather than concentrating it during recommendation events.