Research On Network Water Army Behavior In The Field Of Electronic Commerce

With the rapid development of network,people's daily life has become highly informationized.In order to win the information war,the network Navy came into being.The existence of the navy has seriously polluted the network environment and affected the network security.They rely on the release of a large number of garbage information in a short period of time to seek benefits,which is harmful.Because e-commerce Navy releases garbage information directly to products and other e-commerce areas,it is easy to cause people's property losses.Therefore,the identification research of e-commerce network Navy deserves more attention.First,in order to maximize the benefits in the shortest time and maximize the impact,naval users usually give a unified score on the target commodity,which is usually very high or very low,so that the average score of the commodity can be changed to the greatest extent.Secondly,in terms of the content of comments on commodities,due to the huge amount of work that the Navy needs to complete,it is doomed that he can not write different real feelings according to the differences of different products like normal users,so there will be a large number of identical or similar comments.Thirdly,the evaluation information of normal users and Navy users will be different when they cross together,so we can judge the Navy account by the user's score deviation.In addition,due to the wide application of Wechat Public Number in recent years,there will be a phenomenon that the Navy uploads two-dimensional code pictures with Wechat in the Comment Zone of Commodity Pictures,which undoubtedly disturbs the normal order of comment.There are great differences between these naval behavior patterns and ordinary users.We hope that we can start from this research and analyze naval behavior patterns,so as to achieve the purpose of identifying abnormal users.In addition to the above features,we also observed that there are a lot of unrelated comments or garbage reviews for evaluating other products in order to brush users' stars,which are useless for users and need to be identified.In order to solve this problem,we use LDA topic model to judge whether the comment content is centered on the target commodity,and calculate the deviation degree of the comment text to determine the irrelevance between the content and the theme.Finally,this paper analyses the behavior of the Navy from six aspects,and constructs a detection model.They are single product review score detection model,single product review content detection model,product group detection model,score deviation detection model,independent review detection model and LDA topic deviation detection model.According to the importance of each model to allocate weight,the detection model of e-commerce water army is obtained.Finally,the data are integrated to analyze thefeasibility of the model.However,there are still some shortcomings.Because there is no clear data set for marking Navy users,this paper adopts the method of manual marking navy,which may have some subjectivity and error.There is also that the experimental data is not large enough,the speed is not fast enough,and the real e-commerce platform needs to detect the amount of data is far from,still need to continue to improve.