Research On Weibo Emotional Tendency Integrating User Behavior Analysis

According to the 43rd statistical report on the development of Internet in China released by China Internet network information center(CNNIC),by the end of December 2018,the usage rate of microblog has been increasing continuously,reaching 42.3%,which is an increase of 1.4%compared with the end of December 2017.At the same time,according to the latest data of CNNIC,by the end of December 2018,for sina microblog alone,138,253 government agencies have been certified by the platform,and all provinces(autonomous regions),ministries and commissions,as well as administrative units at the county level,public security law,culture and health have opened microblog platforms.At the same time,with the explosive growth of the number of microblog users,some commercial microblogs and personal microblogs have been mixed together,resulting in an increasing number of false and spoof information,which seriously threatens social stability and public security.It is helpful for the construction of a harmonious society to know all kinds of problems people are concerned about in time and identify and deal with them.In order to make good use of the massive social platform information of microblog,tap the potential value of microblog,collect the key information on microblog platform and analyze its tendency,it is necessary to realize the anal-ysis of network public opinion based on microblog information.This paper aims to analyze the emotional tendency of microblog,and studies the contents including microblog information col-lection and preprocessing,social network impact maximization,network microblog emotion anal-ysis integrating user behavior analysis,microblog public opinion analysis and early warning model and prototype system.Generally speaking,it mainly includes the following four aspects:(1)Aiming at the real-time problem existing in the current microblog information col-lection process,an improved crawling strategy based on time characteristics was proposed to improve the efficiency and timeliness of microblog information collectionDue to the characteristics of microblog web pages,the update frequency of different mi-croblog web pages varies greatly,especially when there is an emergency,the time characteristic of microblog information is particularly important,and the more real-time information is more valu-able.Therefore,in the process of crawling microblog data,the appropriate strategy of crawling microblog data can be adopted to accurately crawl the target data and ensure the timeliness of information at the same time.In the existing crawling strategy of microblog web pages,the dedu-plication strategy for repeated web urls can improve the efficiency of webpage crawling.However,in the crawling process,the latest microblog hot news cannot be obtained in time and the latest public opinion dynamics cannot be mastered.In view of this,this paper adopts the improved crawling strategy of microblog web pages.In view of the fast update speed of microblog web pages,it adds time characteristic mark to the URL captured.When the URL is captured again,the time characteristic mark of the URL and the content of microblog web pages are compared.For the same URL marked with different time tags,corre-lation analysis is carried out on the main content to improve the real-time performance of crawlinginformation.The experiment shows that the improved crawling strategy of microblog web page has better real-time information than the existing crawling strategy of microblog web page,and can better reflect the latest changes of public opinion situation,thus winning more time for guiding and controlling the development of online public opinion.(2)Aiming at the problem of maximizing the influence of social network communication represented by microblog,the paper proposes a social network influence maximization model based on memory effect and social enhancement effect to improve the communication effect and efficiencyBecause of the social network represented by microblog,it has great application value in word-of-mouth marketing and viral marketing.Therefore,the issue of social network impact max-imization based on this can be used to select a limited number of initial nodes in social networks to spread public opinion information and commodity information,so as to maximize the number of customers who have purchased commodities,and can also be applied to the accurate analysis and supervision of public opinion information.Although the existing research on impact maximi-zation simplifies the complexity of the problem,it ignores the dynamic nature of social networks,which is a common phenomenon in social networks and viral marketing networks.This leads to the divergence between practical application and theoretical research and reflects the practical value of theoretical research.Based on this,this paper expands the traditional impact maximization problem into a more practical viral marketing problem based on memory effect and social enhancement effect.Firstly,a subordinate viral marketing model is proposed to capture the dynamics of memory effect and social enhancement effect.Secondly,this paper proposes a new algorithm to solve the problem based on the slave cascade viral marketing model,and verifies the effectiveness of the new algo-rithm through theoretical proof and experimental analysis.Finally,the algorithm model proposed in this paper explains the reason why the actual effect of celebrity microblog marketing fails to reach the predicted result of simulation experiment.(3)Analyzed the main user behaviors involved in the process of microblog communi-cation,and propose an analysis method of microblog emotional tendency integrating user behaviors to improve the accuracy of microblog emotional analysisPeople can express their opinions and communicate with others in social networks through a series of user behaviors such as Posting and forwarding,and researchers focus on how to deal with the information generated by their behaviors on social networks.In public opinion analysis,user behavior can be integrated into microblog emotion analysis and public opinion analysis by ana-lyzing data obtained from user behavior monitoring,so as to improve the effect of microblog emotion analysis and public opinion analysis.Based on this,this paper proposes an emotional analysis scheme for user forwarding mi-croblog based on LDA model.Firstly,emotion labeling is carried out through the constructed emotion dictionary,and the user'^historical microblog and the forwarded microblog are extracted and put into the LDA model through feature extraction.Through the feature representation of theLDA model,the subject probability vector features of the user's historical microblog and the for-warded microblog are obtained.Then calculate the emotional tendency of microblog content and forwarded content.The emotional value of the content forwarded by microblog is calculated by relation weight.Finally,the emotional analysis of forwarding behavior is carried out by combining the emotional tendency,the user's emotional degree to the microblog and the emotional value of forwarding the microblog content.(4)A Hadoop based network public opinion analysis model was designed,and a mi?croblog public opinion analysis and early warning prototype system was implementedHow to quickly mine valuable public opinion information,quickly locate it,track and monitor hot public opinions in real time,and effectively control the spread and development of online public opinions is one of the urgent problems to be discussed and solved in the current supervision of online public opinions in the context of big data.Based on this,this paper proposes an analysis model of network public opinion based on Hadoop,and USES HDFS file service system to store massive network data in a distributed man-ner,providing fault tolerance and reliability assurance.In the process of clustering,aiming at the low efficiency of traditional k-means clustering method in dealing with mass data,this paper adopts the k-means distributed topic clustering calculation method based on MapReduce to effi-ciently deal with mass public opinion information through multi-machine collaboration.Through the topic heat analysis,the hot public opinion information in a certain period on the network is obtained,and the effectiveness of the method in this paper is verified by experiments.Finally,this paper also designed and implemented a network public opinion analysis and supervision prototype system to serve the government management and decision-making.