US Patent 10986408 Systems and methods for recommending linear content

A device receives content data associated with a collection of content that is available to a user. The device determines, by processing the content data using a first data model that has been trained using machine learning, measures of likelihoods that the user will be receptive to each respective content of the collection of content. The device determines a duration during which the user is predicted to engage in a content watching session by using a second data model to process at least a portion of the content data. The device determines measures of utility that correspond to the collection of content and that measure utility that each respective content is predicted to have to the user during the content watching session. The device determines content recommendations for the user based on the measures of utility and causes at least one content recommendation to be displayed.