Deep heterogeneous graph embedding model based on feature fusion

Abstract

The invention discloses a deep heterogeneous graph embedding method based on feature fusion, and the method comprises the steps: carrying out the message propagation between nodes through a graph attention mechanism after a meta-path sub-graph is extracted, aggregating the information of neighborhood nodes, and updating the embedding representation of a central node; multi-head attention enhancement feature learning is used, and a residual connection enhancement model is used to capture local information; aggregating node embedding obtained by attention convolution learning of each layer of residual image by utilizing jump connection; carrying out residual image attention node embedding and interlayer feature fusion on the meta-path sub-graph, and learning an embedded vector representing node information in a single dimension; aggregating node embedding information learned by different semantics by using a fusion function to obtain final node embedding; embedding and projecting the learned nodes into a label category space by using a full connection layer; using a loss function to measure the loss of the predicted value and the true value, and optimizing the parameter updating gradient until the model converges. According to the method, the features learned at different levels can be effectively fused, and the nodes can adaptively select information.

Description

technical field [0001] The invention relates to the technical field of graph embedding, in particular to a feature fusion-based deep heterogeneous graph embedding model and a deep heterogeneous graph embedding method based on the feature fusion deep heterogeneous graph embedding model. Background technique [0002] There are a large number of complex information networks in the real world, such as business networks, social networks, citation networks, biological networks, etc., which contain a lot of valuable information. For example, in a commercial network composed of massive e-commerce data, it contains a wealth of commodity information, transaction information, user behavior information, etc., which contain high commercial value. Data mining of this information in the commercial network can not only realize user intention recommendation to bring huge commercial benefits, but also identify illegal transactions such as cash-out user detection to prevent and resolve financi...

AI Technical Summary

Problems solved by technology

However, the existing heterogeneous graph models lack the careful design of the deep structure, and do not fully express the characteristics of different levels.

Images