The implicit favourites rule allows to detect programs which a user frequently consumes and thus probably likes, even though they might not explicitly have been marked as favourite programs. A given program is marked as an "implicit favourite" of a user if they consumed more than a certain threshold of episodes of that program. Recommended episodes that belong to an "implicitly favourited" program are upscored and therefore more likely to be recommended.
To determine implicit favourites for all users, all watch/listen events of all episodes are periodically collected and grouped by user-program combinations. Each combination passing the threshold is then saved as an "implicit favourite"-association between a user and a program.
The code of implicit favourites computation is available in PEACH Notebooks:
pipe_algorithms > notebooks > tutorials > business-rules > implicit_favourites_task.ipynb. It is generic and can easily be adapted to compute implicit favourites for any content catalogue and consumption data from any organisation. Note that in addition to consumption events, the computation requires metadata that maps episodes onto the programs they belong to.
During each recommendation request, the implicit favourites of the requesting user are looked up, and if there are any items in the preliminary recommendation set belonging to any of the implicitly favourite programs, they get upscored by a fixed factor, potentially resulting in a new ranking of the recommended items.