Prediction device and prediction method for state of emergency topic

Abstract

The invention relates to a prediction device and a prediction method for the state of an emergency topic. The prediction device and the prediction method are provided based on the research of a microblog topic production mechanism and the influence of opinion leaders of emergency on the state of the emergency topic. The prediction device is provided with four components of a text acquisition unit, a topic clustering unit, a topic state recognition unit and a topic state prediction unit, and the topic state is predicted through the participation state of the opinion leaders, so that the prediction method of the topic state is simplified, the prediction accuracy is higher, and a basis can be provided for monitoring the tendency of the emergency topic. Through the calculation method of topic attention provided by the invention, the computation complexity of a traditional topic attention method is reduced. Moreover, the numerical value prediction of the topic attention is replaced by adopting the prediction of the state of the topic, so that the prediction content is more reasonable, and the prediction accuracy is higher. Furthermore, a prediction method of a hidden Markov model is provided, the set of the opinion leaders and the model of topic prediction are constructed through incremental iteration, so that the prediction operation can be obviously simplified, and the prediction efficiency is improved.

Description

technical field [0001] The present invention relates to a topic detection and tracking TDT (Topic Detection and Tracking) technology of the Internet, to be more precise, it relates to a prediction device and method for predicting the topic status of emergencies, which are used to resolve the occurrence of emergencies, micro Identify and track the status of topics in the blog network, and predict the topic status of emergencies through the status of the participation rate of opinion leaders, establish a prediction model, and provide a basis for monitoring the trend of topics of emergencies. It belongs to the technical field of Internet information application. Background technique [0002] Internet topic detection and tracking TDT technology originated from the early event-oriented detection and tracking EDT (Event Detection and Tracking). TDT is oriented to news reports in multilingual text and language forms, and is mainly engaged in related tasks such as automatic identif...

AI Technical Summary

Problems solved by technology

[0013] (1) Information such as topic popularity of emergencies changes irregularly, and the data fluctuates greatly, which can only be kept stable in a small range. Therefore, the set regression model, time series model, etc. require a large amount of prior data. prediction methods, and, also not well suited for topic prediction

[0014] (2) It is not easy to define the demarcation point of the topic life cycle, so depending on the topic, it is impossible to use a unified value to describe information such as topic popularity

Images