GENERATING DIGITAL MEDIA CLUSTERS CORRESPONDING TO PREDICTED DISTRIBUTION CLASSES FROM A REPOSITORY OF DIGITAL MEDIA BASED ON NETWORK DISTRIBUTION HISTORY

The present disclosure relates to systems, non-transitory computer-readable media, and methods for generating accurate digital media clusters corresponding to predicted distribution classes from a repository of digital media based on network distribution history. For example, a digital media clustering system can apply machine learning models at a remote server (based on network distribution history of a network account of a user) to generate predicted distribution classes for future electronic communications. The remote server can provide the predicted distribution classes to a user client device for secure local analysis of digital media stored at the client device. Based on the predicted distribution classes and the stored digital media, the client device can suggest digital media items to distribute via a networking system. Thus, the disclosed system can surface digital media content without providing any information regarding the digital media items stored at the client device to a remote server.