Mining And Derivation Analysis Of Opinions Of Groups Concerned About Specific Events On Weibo

With the emergence and rapid development of social media on the Internet,microblogs provides a convenient space for the accumulation and dissemination of public opinion,which takes user interaction as the core.As a result,many government agencies,media and celebrities have opened microblogs to interact with netizens,and government agencies have used microblogs to solicit public opinions;celebrities have opened microblogs to publish positive energy information for more support;news media published news messages in order to expand awareness.However,there are also many false news and negative news spread on the internet,misleading users' opinions,make negative impact on social stability,national unity and national security.Therefore,identifying focus group at a specific event,and focus group to carry out excavation and opinions derived analysis not only has important significance of research,but also has important practical significance for network monitoring public opinion,listen to the demands of the people,maintain social stability and safeguard national security.Taking Sina microblogs as the research object,this paper studies the identification of specific event focus group,focus group's opinions mining and analysis of the key elements of opinions evolution.The research work covers the construction of distributed data computing platform,specific event focus group identification,opinion mining,sentiment analysis and opinion evolution analysis.Forming a set of methods and theories of microblog specific event focus group's opinions mining and evolution analysis.The main research work and results are as follows:(1)Research on the acquisition of microblog data and the construction method of data computing platformThis paper propose a microblog data acquisition method and a data preprocessing strategy.Through microblog API and Web crawler based on Python to collect specific event data,the database model is constructed according to the data format,and then layer the data.It is divided into ODS layer,PDW layer and APP layer.At the same time,propose the strategy of constructing distributed data computing platform.By analyzing and selecting the data calculation framework,Hadoop+Hive+Spark+Mysql batch computing platform and Storm streaming data computing platform were constructed to calculate and analyze massive microblog data.(2)Research on the method of focus group identification for specific eventsA focus user activity calculation model of specific events based on static features and dynamic attributes is proposed.By extracting user attributes features and dynamic interaction attribute features to build a specific event focusing on user activity model.The weight value of user activity parameter is calculated by analytic hierarchy process(AHP).At the same time,a user mining algorithm based on network structure features is proposed.Based on the PageRank algorithm,the user attribute weight and the user interaction attribute weight feature are added,and the DLRank key user recognition model is constructed.Finally,focus on the users for feature extraction,and the feature attribute vector representation strategy based on feature mapping is constructed.The c-means clustering algorithm is used to focus group of specific events.The experimental results show that proposed model can identify the focus users of specific events,at the same time,the feature mapping method is used to cluster the user groups,and the average clustering accuracy is at least 0.06 percentage points higher than the hard matching rule.(3)Research on the focus group's opinions Mining Method Based on Sentiment AnalysisIn this paper,a focus group's opinions mining model based on sentiment analysis is proposed.Firstly,construct a sentiment dictionary for the microblog field to improve the accuracy of NLPIR word segmentation,and then use Standford Parser for syntactic analysis to extract the dependencies in the microblog data.At the same time,the subjective sentence recognition model based on maximum entropy is proposed.The subjective sentence recognition is applied to the microblog data,then calculated the sentiment tendency probability of the subjective sentence,and the sentiment tendency with the largest probability value is selected as the sentiment polarity of the user.For the calculation of the parameter weight value in the maximum entropy model,an iterative scaling algorithm is used for training.The experimental results show that the proposed model can effectively identify the subjective sentences in the microblog text,and the F value of the sentiment orientation of subjective sentences is more than 80%,which proves that the model constructed in this paper is feasible.(4)Research on the focus group's opinions Mining Method based on the Topic AnalysisThis paper proposes a focus group's opinions mining model based on topic analysis.Improves the TF-IDF algorithm,added the part of speech feature and constructed the POS_TFIDF subject extraction model.At the same time,based on the POS_TFIDF model,the mutual information of co-occurring words and the location information characteristics of co-occurring words are considered comprehensively,and the MLCP_TFIDF subject extraction model is constructed.Mining the perspectives of specific event focus groups based on the above two models.The experimental results show that the topic mining model proposed in this paper can effectively extract the topic keywords of the focus group,which has certain practicability and feasibility.(5)Research on Identifying Key Elements of opinions evolution in Specific EventsA focus group's opinions evolution analysis model based on sliding time window is proposed.Firstly,according to the development track of a specific event,the specific events are divided into time windows,and the specific events in the adjacent time window are used for opinion mining.Then,the similarity degree of the opinions in the adjacent time window is calculated,and the opinions evolution trajectory is analyzed according to the opinion evolution judgment rule proposed in this paper.If the opinions is evolutioned,the time window of evolution and the characters,time,place,key subject words and other factors in the adjacent time window are extracted and analyzed,and the key elements affecting the evolution of the focus group are identified.The experimental results show that the model proposed in this paper can effectively monitor the trajectory of the focus group,and identify the development of the model in a timely manner.At the same time,the key factors that affect the development of the focus group's opinions are relatively accurate.And the key factor of the influence point of opinions can be effectively identified.Finally,a micro-blog specific event focus group's opinion mining and evolution analysis system is constructed.The system consists of microblog data collection and preprocessing,specific event focus group recognition,specific event focus group's opinion mining and specific event focus group's opinion evolution analysis.It can perform experiments and visual display on related algorithms such as data collection,opinion mining,sentiment analysis and opinions evolution for specific events.