Sales Performance Predicting By Using Online Reviews Of Textual Polarity Sentiment

With the continuous development of Internet technology,network public opinion has become a heated topic and arouse social concern.Among the topic,sales performance prediction with textual sentiment analysis of online reviews is becoming a hotspot.Sales performance prediction with textual sentiment analysis of online reviews not only scientifically analyze users' sentiment attitudes towards products of online reviews,but also have important commercial value for online enterprises to predict product sales.Based on the research background and significance of online reviews,this paper first introduces the concepts and theories,combs the relevant literature at home and abroad,and analyses the research status of online reviews,which include the impact of online reviews on sales,the extraction of emotional information from online reviews,and the trend prediction based on online reviews.Secondly,after summarying the existing models(classical sentiment model and mainstream prediction model),this paper propose the DS-LDA model and MDSA model based on the innovation and optimization of traditional models.Thirdly,we use Octopus Crawler software to capture the online reviews of 30 IMDB movie websites,focusing on sentiment polarity feature extraction and separation of online reviews,and exploring the use of these features and eliminating periodic box office data to predict box office revenue.Finally,through a large number of experimental data,the model constructed in this paper is evaluated repeatedly to obtain the optimized model,DS-LDA model and MDSA model.The contributions of this paper are mainly reflected in two aspects:(1)The construction of polarity sentiment analysis model.Compared with the traditional model,this paper will be judged by sentiment polarity or extract sentiment features.This model can extract and separate sentiment polarity features from online reviews,and further understand users' sentiment attitudes toward products.(2)The construction of the manifold dynamic sentiment aware model.Compared with the traditional prediction model which needs pre-processing to remove theperiodicity,this model can directly remove the periodicity of product sales,and integrate unstructured text data and structured product sales data.It can effectively solve the problem of periodicity and data fusion on product sales.The experimental results show that the model constructed in this paper has higher accuracy and effectiveness.