Recurrent neural network-based social network message burst detection method and system

Abstract

The invention discloses a recurrent neural network (RNN)-based social network message bust detection method and system, and relates to the technical field of popularity prediction of contents in social networks. The method comprises the following steps: acquiring history messages published and forwarded by a user in a social network, and preprocessing the history messages to obtain a history forwarding time sequence; carrying recurrent neural network training on the history messages and the history forwarding time sequence, and generating a prediction model; and acquiring messages published and forwarded by the user in real time, generating a forwarding time sequence according to the messages, inputting the forwarding time sequence into the prediction model to generate feature expressions, inputting the feature expressions into a fully-connected neural network to carry out classification, and outputting the classification result in a softmax manner so as to complete the social network message burst detection.

Description

technical field [0001] The present invention relates to the technical field of popularity prediction of content in online social networks, in particular to a method and system for detecting outbreaks of social network messages based on a recurrent neural network. Background technique [0002] Online social media represented by Weibo generates hundreds of millions of news and content every day. The highly interconnected structure of users in social networks and the herd effect of users make the spread of news very convenient and efficient, which greatly facilitates It improves the way people obtain information and strengthens the connection between people. However, the information in the network is uneven, and only a very small part of the news will eventually become popular and erupt, causing widespread public concern and accompanied by huge public opinion and Impact, effective identification at the early stage of news or content outbreak is an important means for online rep...

AI Technical Summary

Problems solved by technology

[0003] At present, predicting the outbreak and popularity of news can be mainly divided into analysis methods based on content characteristics and methods based on self-motivation point process. The method based on content characteristics detects sudden changes by capturing the abnormal changes of news-related content characteristics over time. topic, this method requires the accumulation of news dissemination to a certain extent, reaching a significant level, which objectively causes the time when the outbreak news is detected to be close to or lags behind the time when the news outbreak actually occurred, and the timeliness is not high; based on self- The method of stimulating the point process takes the message individual as the object, and models its forwarding time series as a self-motivated point process, aiming to describe the sequence characteristics of "the rich get richer" and "time decay" in the message dissemination, and based on Compared with the content feature method, this method has high timeliness, but the method based on the self-motivation point process still has the following disadvantages: First, its features are artificially defined and strongly dependent on data. For the modeling of the phenomenon of "getting richer", some use linear functions, and some use nonlinear functions. For the "time decay" effect, some use log-normal distribution, and some use power law distribution; second, the model predicts When using only the observation sequence of the news to be predicted, the historical dissemination information of other news is not used, resulting in low prediction accuracy

In summary, there is still a lack of a timely prediction, less human intervention and a good prediction effect method

Images