Abstract: A device identifies, by using a first model to process a first weighted feature set, users who are predicted to stop using a service provider for one or more services. The device determines, by using a second model to process a second weighted feature set, user scores for the users who represent predicted value that the users provide to the service provider. The device identifies, based on the user scores, particular users, of the users, as targets to be offered additional services. The device determines, by using a third model to process a third weighted feature set, service scores that represent predicted levels of interest of a user, of the particular users, in a set of services. The device selects services based on the service scores. The device causes a services package that includes the services to be provided to a user device or an account associated with the user.