Application Of Sentiment Analysis Of User Reviews Supporting Internet Marketing Activities

With the rapid development of the Internet,more and more users comment on the Internet marketing platform or live broadcast platform.The move of users to comment on the Internet makes the amount of text information grow exponentially.These huge numbers of comments attract the attention of enterprises and turn the focus to the emotional analysis of user comments.Through the analysis of the users’ comments data of an Internet marketing platform in Zhejiang Province,this paper analyzes the emotional tendency expressed by them,and deeply analyzes the users’ views on them.It can be found that the main problems that users are concerned about are product quality and enterprise service.The theme analysis of quality and service is carried out respectively.Combined with the actual situation,in order to help enterprises understand more about the Internet product reputation and the rapid development of enterprises Accurate insight into the market competition,put forward suggestions to solve the problem.This paper mainly discusses from the following aspects:1)The collection and process of users’ comment data.The data of this paper was derived from the user comment of a We Chat official account.The data is abundant and rich in content.But due to the limited technological conditions,the data collected are jumbled and confused.The corresponding models of text sentiment classification require higher generalization ability.Therefore,it is necessary to perform preprocessing operations such as data integrity and data transformation,which is an important basis for sentiment analysis.In this paper,it mainly describes the collection of training corpus related to user comment sentiment analysis,as well as the process of its preprocessing.2)Naive Bayes probability problem.Naive Bayes has excellent performance in emotion classification,but the algorithm still faces two problems,one of them is that the attributes need to be independent of each other,the other is the method of rough probability estimation.It finally produces a new algorithm which can improve probability estimation is developed by optimizing and transforming based on Naive Bayes.The probability optimization function enables Naive Bayes to fully consider the situation that the probability of the condition is equal to zero,thus,avoiding many problems,such as overfitting and easily underflow.In this paper,10 datasets in UCI are selected from a huge number of datasets,and on this basis,several groups of comparative experiments are carried out.The experimental results show that the proposed method performs better in the classification accuracy of noisy and large datasets,which shows the effectiveness of the method compared with the existing methods,and also provides conditions for subsequent experiments.3)Research on fusion of Naive Bayes and Decision Tree algorithm.In this paper,a new algorithm is proposed,which is the fusion of Naive Bayes and decision tree.In the process of training,there could be noise contradictions in some individual cases,and the Decision Tree will be overfitted,which will influence its accuracy.The improved probability optimization algorithm based on Naive Bayes can remove the interference noise in the training set,and construct the Decision Tree on this basis,which can solve the problem of over fitting in the decision tree.This paper uses the data set of user comments of We Chat official account to conduct experiments on the algorithm proposed in the paper.Finally,the data set is divided into two categories: positive and negative,and then the visual data graph of positive and negative comments is achieved by constructing data semantic network.4)Find and solve the problems of Internet products.Based on the user reviews of an Internet marketing platform in Zhejiang Province,this paper uses clustering algorithm for clustering analysis;Compared with the clustering results,it is found that users pay most attention to the quality and service of Internet products,and follow the principle of the most times.Mining the characteristics of quality and service from the data samples,it is found that users care most about the interface layout,usability and product function for the quality,and most about the after-sales service,product activities and user experience for the service;Finally,this paper puts forward some methods that can be applied to point out the direction for the future development of Internet products.