Sentiment Analysis Research And Competitive Intelligence Mining Application Based On Added Expression Features

Currently,a large amount of user-generated content provided by online media such as blogs,online communities,and social networking sites can be used to mine business competitive intelligence.Companies have been researching consumer demand and urgently need to identify consumer opinions and competitive strengths and weaknesses in a highly competitive market.Analyzing consumers' perceptions of their products and competitors can help decision makers gain insight into the dynamics of human and social behavior behind competition.Researchers generally obtain data through Taobao.com and Jingdong websites.These comments are mostly based on praise.Due to various factors,these favorable rates are not true and will cover the defects of the products.Therefore,this article will select the source of data from China's most popular social networking platform-Sina Weibo,in Weibo people will publish a true view of some products and user experience.The research subjects were selected as the top three Apple,Huawei and OPPO brands in China's mobile phone market.Text sentiment analysis is divided into two methods based on dictionary and machine learning.This paper chooses the sentiment analysis method based on machine learning.The Octopus data collector is used to obtain the evaluation contents of the three major mobile phone brands by the Weibo users,and the collected corpus is cleaned and deduplicated,and the emotional classification labels are manually labeled.The data was divided into training sets and test sets in a 7:3 ratio,using R studio as the experimental environment.The corpus is divided into words,feature extraction,vectorized representation,and weight calculation.The information gain IG method is used to select the feature of the text vocabulary,the microblog expression is added to the feature design,the TF-IDF weight calculation method is selected for weight assignment,and the vector space model is applied as the text representation method.The support vector machine(SVM)is used as the support vector machine(SVM).An emotional classification model that trains and predicts data.The accuracy,recall rate and F value are used as indicators for judging the classification performance.The trained classification model is sentimental analysis to the new sample set,and the emotional distribution results are used to analyze the emotional distribution of the product and the attribute details,structural analysis and visualization,mining mobile phone competitive intelligence,analyzing the brand's own advantages and disadvantages and comparing competitors' superiority.Disadvantages are conducive to consumers making purchasing decisions,and also provide reference value for determining the marketing strategy for business strategy development.The innovation of this paper is that the data sources of research selection for intelligence mining are mostly shopping websites,and the data source selected by the article is Sina Weibo website.For the language features of Sina website,Weibo expressions are added to feature design.In the paper,there are two feature selection methods including microblog expression features and without microblog expression features.The results show that the emotional classification effect of microblog expressions is better,and the emotional preference judgment of mobile phone comments in Weibo can be more accurately.At the same time,the three commonly used sentiment analysis classification methods of support vector machine,random forest and naive Bayes are compared to verify that the support vector machine model is more suitable for Weibo sentiment analysis.