METHOD AND SERVER FOR GENERATING MACHINE LEARNING MODEL Russian patent published in 2024 - IPC G06N20/00 

FIELD: physics.

SUBSTANCE: invention relates to a method and a server for generating digital content. Method includes receiving, by a server, a content recommendation request from an electronic device, wherein the user is associated with user-item interaction data indicating previous interactions of this user and other users with a plurality of digital items available for recommendation in the recommendation system; formation by server, using a trained model of element-dependent decision trees (ISDT), estimates of the probability of interaction of a given user with one or more digital elements from a plurality of digital elements, wherein to form an estimate for a certain digital element in said model, a certain set of ISDT associated with such digital element is used and trained independently of the remaining ISDT sets contained in said model and associated with other digital elements, wherein the following data from said user-element interaction data (240) is used as training data for said defined set of ISDT: training target set for said specific digital element, containing data on user interactions with said specific digital element; training input set for said specific digital element, containing data on user interactions with other digital elements from said plurality of digital elements, wherein data on interactions of other users with a subset of other digital elements from said plurality of digital elements are used as training features for said defined ISDT set, sending a response to the electronic device with instructions as a recommendation of those digital elements from said one or more digital elements with which the generated estimates indicate the highest probability of interaction of the given user.

EFFECT: high efficiency of generating digital content.

9 cl, 6 dwg