Object feature information acquisition method and device, object classification method and device and information pushing method and device

Abstract

The embodiment of the invention provides an object feature information acquisition method and device, object classification method and device and information pushing method and device. When the feature information of an object is determined, a plurality of neighbor nodes of a first node are respectively determined from N relational networks at N moments, so that N neighbor node groups can be obtained; spatial aggregation features of the first node at each moment is determined based on the neighbor node group corresponding to each moment and the node feature of the first node; the N spatial aggregation features at the N moments are inputted into a sequence neural network sequentially according to a time sequence, N spatial-temporal expressions of the first node at the N moments can be obtained respectively; and the N space-time expressions are aggregated, so that the space-time aggregation feature of the first node can be obtained and is adopted as the feature information of a first object represented by the first node.

Description

technical field [0001] One or more embodiments of this specification relate to the technical field of graph computing, and in particular to methods and devices for acquiring object feature information, classifying objects, and pushing information. Background technique [0002] In the era of big data, a large amount of object-relational data can be obtained, through which a relational network of multiple objects can be constructed, and nodes are used to represent objects. For example, a relational network for objects such as users and / or items can be constructed. For the relational network graph, usually, through the graph embedding algorithm, based on the initial characteristics of the nodes and the connection relationship between nodes, the embedding vector of each node in the relational network can be calculated, that is, the deeper characteristic information of the object can be obtained. The feature information of the object is represented by a vector of predetermined d...

AI Technical Summary

Problems solved by technology

However, the relationship network itself is dynamically changing. For example, in the social relationship graph, due to the continuous generation of new friend relationships and the dissolution of friend relationships, the structure of the relationship network at different moments is likely to be different.

Therefore, only using the network structure information at one moment to determine the feature information of each object does not make good use of the dynamic change information of the previous network structure.

Images