Multimedia recommendation method and system based on main graph learning technology

Abstract

The invention discloses a multimedia recommendation method and system based on a main graph learning technology, and belongs to the field of data mining. The recommendation method comprises the following steps: collecting behavior data of a user and corresponding media content data, and establishing a database; constructing a multimedia recommendation model based on a main graph learning technology; using the behavior data of the user to train a recommendation model based on a Bayesian personalized sorting thought; according to the established database, a trained multimedia recommendation model based on the main graph learning technology is used, and media information recommended to each user is output; aiming at the problem that inherent noise and redundant information exist in user behavior data, core behavior intentions of users are mined based on a main graph learning technology, and user behavior characteristics are modeled more effectively; and meanwhile, a self-supervised learning auxiliary task is constructed, so that the extracted user features better fit user preferences, the user features are prevented from being polluted by noise behaviors, and the distinguishability of the user features is further enhanced.

Images