FEATURE CLUSTERING OF USERS, USER CORRELATION DATABASE ACCESS, AND USER INTERFACE GENERATION SYSTEM

摘要

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for a feature clustering of users, user correlation database access, and user interface generation system. The system can obtain information stored in different databases located across geographic regions, and determine unique users from the different information. The information can be included in unique records in the databases, with each record describing a particular user, and with each user described with imperfect identifying information. The system can analyze the different information utilizing machine learning models, and can associate each record with a particular unique user. The system can obtain identifications of items associated with each user, and determine the propensity of the user to disassociate with one or more items, or determine likelihoods of future association with different items not presently associated with the user.

主权项

1. A computerized method performed by one or more computer systems, wherein the computer systems are configured to access one or more electronic data sources in response to requests received from an interactive user interface, the method comprising:
obtaining, from the one or more electronic data sources each storing user records, information describing a plurality of users; determining, using the obtained information, a plurality of unique users and user records associated with the unique users; obtaining, for each unique user and from associated user records, identifications of one or more services connected with the unique user and information describing one or more features of the unique user; determining, for each unique user, information indicating likelihoods of the unique user disconnecting from one or more services connected with the unique user; determining, for each unique user, one or more recommended services for the unique user comprising:
determining clusters to assign each unique user according to one or more features of respective unique users,obtaining identifications of services connected with unique users that are included in a same cluster as the unique user,determining one or more of the obtained services that the unique user is not connected with, andassigning the one or more determined services as the one or more recommended services for the unique user; and providing, for presentation in an interactive user interface, summary data describing one or more unique users.