A social media user account classification method based on information fusion

Abstract

The invention discloses a social media user account classification method based on information fusion, comprising the following steps: S1, input social media data; S2, select seed users from social media data, and obtain text information of seed users; S3, perform text Preprocessing, extracting the required node information; S4, constructing a heterogeneous information network, merging the extracted node information to obtain a feature vector; S5, using the heterogeneous information network to construct a graph convolution attention network, and analyzing social media user accounts Classification. The present invention integrates different types of node information by building a heterogeneous information network, restores the complex state of the real network, enriches network information, and finds potential relationships between accounts; and builds a heterogeneous information network based on the heterogeneous information network The quality map convolutional network and the attention mechanism are added to obtain the importance of the influence of different types of nodes in the network on the node, and to improve the accuracy of account classification.

Description

technical field [0001] The invention relates to a social media user account classification method based on information fusion. Background technique [0002] With the rapid development of Internet technology, social media, as a product of the Internet age, has become an indispensable part of people's lives. User accounts, as publishers and disseminators of information, contain a large amount of valuable data information. Therefore, the targeted identification and classification of massive accounts will help reduce the human resources and time costs required for the construction of traditional account management systems, and also enable more comprehensive and effective acquisition of real-time information and dynamics in a certain field. [0003] The existing account classification technology mainly consists of the following two methods: account classification method based on machine learning and account classification method based on deep learning. [0004] Account classific...

AI Technical Summary

Problems solved by technology

However, this method is only suitable for homogeneous networks, that is, the nodes are all of the same type of network, and due to the complexity of social networks, it is not enough to only use homogeneous networks

[0009] In terms of information utilization, the existing account classification methods usually only consider one or more types of information separately, such as the basic attribute characteristics or text features of the account, and do not consider the connection between different information.

At the same time, due to the inaccuracy, incompleteness, and ambiguity of social media data itself, the existing account classification methods cannot obtain accurate account classification results.

Images