Cluster-Level Vector Similarity for Large Data Sets

The patent segments the user profiles into clusters based on shared characteristics and features. Instead of comparing every user profile against every other user profile (user-to-user), the system groups similar users into clusters and performs similarity detection at the cluster level. This segmentation dramatically reduces the computational complexity while maintaining the ability to identify similar user profiles, as the cluster representations capture the essential similarity patterns.

The patent creates simplified representations (copies) of user profiles in the form of cluster vectors that capture the essential characteristics of groups of users. Instead of processing the full complexity of individual user profiles with all their features, the system uses these cluster-level vector representations to perform similarity comparisons. This copying approach allows the system to handle large data sets efficiently by working with condensed representations that retain the necessary information for similarity detection.

The patent performs feature selection and clustering operations in advance to create pre-computed cluster representations and vectors. By performing these computationally intensive operations beforehand (preliminary action), the system prepares the data in a format that enables efficient similarity comparisons later. This preliminary processing reduces the computational burden during actual similarity detection, as the heavy lifting of feature selection and clustering has already been completed, leaving only straightforward vector comparisons.