Heterogeneous graph classification method based on double-layer attention mechanism

Abstract

The invention discloses a heterogeneous graph classification method based on a double-layer attention mechanism, and the method comprises the steps: S11, carrying out the mapping operation of the feature vectors of different types of nodes, and obtaining the feature vectors of a target node and a neighborhood node corresponding to the target node in the same entity space, S12, learning weights ofdifferent types of neighbors and different adjacent nodes from top to bottom from the type-level attention to the node-level attention according to the obtained feature vectors and based on the type-level attention and the node-level attention in the double-layer attention, S13, constructing a heterogeneous graph according to the obtained weight to obtain a classification model, and S14, inputtingtest data into the obtained classification model, and outputting a final classification result.

Description

technical field [0001] The invention relates to the technical field of target customer classification, in particular to a heterogeneous graph classification method based on a two-layer attention mechanism. Background technique [0002] In business, companies will adopt some marketing methods to promote their products. Since different types of products target different customers, they usually classify the target customers of different types of products, and formulate corresponding marketing plans according to the categories of customers, which is conducive to mining. Potential customer base, realize precise marketing, and increase revenue. In traditional methods, it takes a lot of time and manpower to realize the classification of target customers. With the continuous development of Internet technology, deep learning technology can be applied to the above business scenarios, greatly reducing costs. [0003] Most of the data in the real world exists in the form of graphs, su...

AI Technical Summary

Problems solved by technology

The above frameworks are only applicable to heterogeneous graphs with meta-paths. Since meta-paths need to be manually formulated, and the semantic information in heterogeneous graphs is difficult to exhaust through multiple meta-paths, some nodes will be ignored when looking for the neighborhood nodes of the target node. Link edge types or node attributes, which inevitably lead to information loss

Therefore, how to learn representations for heterogeneous graphs without losing heterogeneous information is challenging.

HetSANN directly uses the structural information of heterogeneous graphs for representation learning, abandoning the meta-path in the traditional method, but it only considers the difference of influence of different nodes in the process of neighborhood aggregation, and does not consider the type information of nodes

[0005] The complexity and heterogeneity of the nodes and connections in the heterogeneous graph make the classification method difficult. It is impossible to directly use the classification method of the homogeneous graph to find and classify the target nodes in the graph, otherwise some information in the graph will be lost.

And the meta-path in the traditional method inevitably brings information loss

Images