Construction Of Microblog User Group Classification Model Based On Sentiment Analysis

Analyze the distribution of user groups in Microblog can help us quickly understand the user composition and sentimental tendencies of Microblog topics,and help companies understand their user groups to recommend products more targeted,help the government to control the trend of public opinion in a timely manner.Aiming at the problem that the existing research only considers the user's own attributes or sentimental tendencies and cannot accurately classify the user groups with different opinions under the topic,this paper constructs a user classification model from the perspective of the Microblog text published by the user and text sentiment under the topic.The model classify user groups by the user characteristics extracted from the user's sentimental opinion representation vector,thereby improving the accuracy of user group classification.The main research contents are as follows:(1)The construction of the Microblog sentiment vocabulary and affective computing rules.The Microblog sentiment vocabulary constructed in this article contains two parts:a basic sentiment dictionary and a modifier dictionary.The basic sentiment dictionary includes the sentiment vocabulary ontology database of Dalian University of Technology and the sentiment word dictionary of Internet terms.The modifier dictionary includes a negative word dictionary and a degree adverb dictionary.When formulating sentiment calculation rules,the influence of sentence patterns and the relationship between sentences on sentimental expression are taken into account in this paper.and formulates inter-sentence sentiment calculation rules and sentence pattern sentiment calculation rules.(2)Extract user characteristics based on user sentiment view representation vector.When extracting user classification features,this article innovatively takes into account the user's comment text posted under the topic and sentimental tendencies.First,by introducing word vector technology and clustering algorithm,select the semantic center word of the word as the text feature word.Then,when calculating the weight of feature words,introduce the word similarity coefficient and the sentimental tendency intensity coefficient to avoid the loss of text features and improve the accuracy of the expression of the user's opinion and sentiment text vector.Finally,construct the user's sentimental opinion representation vector based on the text feature words and their weights(3)Constructing a user classification model under Microblog topics based on user characteristics.User characteristics extracted by establishing user opinion representation vector The clustering algorithm k-means is used to construct a user classification model under the Microblog topic.Use user sentiment opinion representation vector as algorithm input,classify the user group based on the user's sentimental opinion feature.When classify user groups,firstly,calculate the contour coefficient SSE to determine the optimal user group division.Finally,the user group classification results under the topic are obtained according to the user's opinions and attitudes on the Microblog topic.The experimental results show that the user classification model under Microblog topics based on user opinions and sentiment can classify user groups accurately and effectively.This paper innovatively introduces user sentiment and textual opinions as features to classify users under Microblog topics.The word similarity calculated has been introduced and the semantic center word is determined as the feature item when the user sentiment feature vector has been constructed.On the basis of the TF-IDF method to calculate weights,the word similarity coefficient and sentimental tendency factors are innovatively introduced to comprehensively calculate the weights of feature words when the weights of feature items have been calculated.This method above can improve the accuracy of the vector of user opinions and sentiment expression and the results of the user groups classification under Microblog topics.The model constructed in this paper can classify user groups under hot topics effectively and accurately.Figure [26] Table [17] Reference [74]]...