Link prediction method based on extensible representation of dynamic heterogeneous information network

Abstract

The invention belongs to the field of data analysis, and discloses a link prediction method based on extensible representation of a dynamic heterogeneous information network, which comprises the following steps of: obtaining scientific cooperative dynamic heterogeneous information network data including network node and network edge data; an embedding mechanism in a complex space is introduced to represent a given dynamic heterogeneous information network at a timestamp 1; learning a dynamic heterogeneous information network from a timestamp 2 to a timestamp t by adopting a ternary graph dynamic embedding mechanism; processing a heterogeneous information network from a timestamp 1 to a timestamp t by using a deep automatic encoder based on a long short-term memory network, and performing graph prediction of a timestamp t + 1 after analysis and calculation; and performing link prediction on the nodes in the network by using the graph data from 1 to t + 1 to obtain a prediction result. According to the method, training is carried out on a change data set based on a meta-graph mechanism, a large-scale dynamic heterogeneous information network can be expanded, and a future network structure can be predicted.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, in particular to a link prediction method based on scalable representation of dynamic heterogeneous information networks. Background technique [0002] Content representation is a fundamental task in information retrieval. The purpose of representation learning is to capture the features of informative objects in a low-dimensional space. Most studies on representation learning for Heterogeneous Information Networks (HINs) focus on static HINs. In reality, however, the web is dynamic and constantly changing. A Heterogeneous Information Network (HIN) is an evolving network with multiple types of nodes and edges. In fact, most networks are dynamic heterogeneous information networks, such as social networks and bibliographic networks. Thus, compared to static networks, dynamic heterogeneous information networks are a more expressive tool for modeling information-rich problems. ...

AI Technical Summary

Problems solved by technology

This makes most existing static embedding models that need to process the entire network step by step, inappropriate and inefficient

Images