Research On Hot Topic Tracking And Relationship Detection Based On Parallel Association Rules

With the large-scale popularization and rapid development of the Internet,more and more people tend to express their attitudes and opinions in Internet,a large number of new media platforms have been spawned to speed up the explosion of network data.Therefore,it is an important issue in public opinion analysis to deeply mine the value of network data to support decision.As a research direction in this field,topic detection and tracking has been widely developed.But the current research has some defects and gaps in using the potential relationship of keywords to track the evolution trend and propagate path of hot event and find the relationship of multiple hot topics.Basis of the deep analysis of the traditional association rules,this article made the following exploration.Since most of the existing methods calculate the semantic distance or the semantic similarity of feature keywords in the continuous time slice to track hot events,it is short of the in-depth analysis on dynamic evolution and propagation path of hot event by using the relationship between the keywords.For this problem,the evolution and propagation path tracking of hot event based on parallel association rules was proposed.Firstly,the frequent keyword sets of multiple time window were obtained by implementing association rules in parallel;Secondly,the association rule sets were obtained by selecting all of the association rule of frequent keyword sets,and then got the keyword sets of multiple topics;Finally,the relationship between the dataset of adjacent time window was found according to the keyword sets,so as to analyze the evolution and propagation path of multiple topics.Experiments showed that our method could track the specific change details of the emergence,development and demise of hot event more comprehensively and effectively,which provided reference and support for network public opinion on the monitoring and management.Since the hidden relationship between topics is not always intuitive,it is possible to look for the deep or root causes of the specific event by finding the relationship of multiple topics in the complex network public opinion data.This paper intro duced the concept of "related keyword set" and proposed a method of topic relationship detection based on parallel association rules.Firstly,the large-scale frequent keyword sets were obtained;Secondly,related keyword sets for each topic were obtained by parallel association rules.Finally,keywords information was obtained by selecting and assembling all of the related keyword sets,so as to find the potential relevance between topics.Experiments showed that our method could find the internal relations of multiple public opinion topics accurately and effectively,which made the strategy more targeted.