Method and device for training interaction prediction model and predicting interactive objects

Abstract

The embodiment of this specification provides a method and device for training and using an interactive prediction model. In this method, firstly, a dynamic interaction graph is constructed based on the sequence of interaction events, from which the first sample nodes belonging to the first type of objects and the candidate nodes belonging to the second type of objects are determined. Using the characterization network, the node vectors of each node are determined respectively. Then, the vector of the first sample node is input into the generation network, and the generation network selects the prediction node from the candidate nodes. Input the first sample node and the prediction node into the discriminant network to determine the first probability of their interaction; also input the first sample node and the corresponding second sample node into the discriminant network to determine the second probability of their interaction . To increase the second probability and reduce the first probability as the goal, train the representation network and the discriminant network; to increase the first probability as the goal, train the representation network and the generation network. The trained representation network and generative network serve as an interaction prediction model for predicting interactive objects.

Description

technical field [0001] One or more embodiments of this specification relate to the field of machine learning, and in particular to training an interaction prediction model, and a method and device for predicting interactive objects using the trained interaction prediction model. Background technique [0002] In many scenarios, user interaction events need to be analyzed and processed. Interaction events are one of the basic elements of Internet events. For example, the click behavior when a user browses a page can be regarded as an interaction event between the user and the content block of the page, and the purchase behavior in e-commerce can be regarded as the interaction event between the user and the product. The interaction event between accounts, and the transfer behavior between accounts is the interaction event between users. A series of user interaction events contains the characteristics of the user's fine-grained habits and preferences, as well as the characteris...

AI Technical Summary

Problems solved by technology

[0003] However, the interaction event involves both parties, and the state of each participant can change dynamically. Therefore, it is very difficult to accurately express the characteristics of the interaction participants by comprehensively considering the various characteristics of the interaction participants.

Images