Methods, systems, and apparatuses for improved content recommendations

A method comprising:

• receiving, by an analytics subsystem of a computing device, activity data indicative of a plurality of engagements of a plurality of user devices with at least one media asset of a plurality of media assets via a client application and a corresponding plurality of user profiles, wherein the client application causes each engagement of the plurality of engagements to be synchronized with a corresponding user profile of the plurality of user profiles;
• updating, based on the activity data, a user interest cloud associated with a first user profile that is not among the plurality of user profiles, wherein the user interest cloud is generated based on the activity data and at least one content feature, of a plurality of content features, and at least one interest attribute, of a plurality of interest attributes, associated with each media asset of the plurality of media assets;
• determining at least one triggering event associated with the client application;
• and
• retraining, based on the at least one triggering event, a trained classification model, wherein the trained classification model is associated with the client application.

A method comprising:

• receiving, by an analytics subsystem of a computing device, activity data indicative of a plurality of interactions of a plurality of user devices with a plurality of media assets via a client application and a corresponding plurality of user profiles, wherein the client application causes each interaction of the plurality of interactions to be synchronized with a corresponding user profile of the plurality of user profiles;
• generating, based on the activity data, a user interest cloud comprising a plurality of content features and corresponding interest attributes associated with the plurality of media assets, wherein the user interest cloud is associated with a first user profile that is not among the plurality of user profiles, wherein the user interest cloud is generated based on at least one content feature of the plurality of content features and at least one interest attribute of the corresponding interest attributes associated with the plurality of media assets;
• determining, by a classification model, based on the user interest cloud associated with the first user profile, a certainty match for at least one content recommendation for the first user profile, wherein the certainty match is indicative of a predicted level of interest for the at least one content recommendation, wherein the classification model is associated with the client application;
• determining, based on the certainty match, that the predicted level of interest does not meet or exceed a threshold level of interest for the at least one content recommendation;
• and
• causing, based on the predicted level of interest not meeting or exceeding the threshold level of interest, a first user device to output at least one fallback content recommendation, wherein the first user device is associated with the first user profile.