Recommended system Architecture _ AI

Recommended system Architecture

The actual recommendation system usually uses a variety of recommendation algorithms, and according to the user's real-time behavior feedback to adjust the user's feature vector (feature weighting coefficient), and then fusion of recommendations of the recommended algorithm, on this basis to filter the recommended items, and finally combined with users to adjust the recommended results rankings, give the final recommendation results.

Based on the different characteristics of the recommended algorithm often use periodic calculation, regularly updated feature items recommended table, such as based on the item similarity characteristics, can save each item the most relevant K item; Based on the user, keep each user the most recent n item, based on the tag feature, Save the maximum number of m item per tag, and save the hottest n item for each age group based on the user's age feature, save the n item that each user likes recently, or a favorite m category based on user like.

The user's real-time behavior feedback and the user's current scene will affect the final recommendation results in real time, and the user's real-time feedback can directly affect the fusion of the recommendation results, and the user's scenario will decide the ranking and presentation of the recommended results. User feedback will also affect the offline computing of the items recommended data.

Reference: "Recommendation System Practice"