Research On Movie Review Spam And Spammer Detection Technology Based On Deep Learning

With the rapid development of the Internet and the film industry in China,online movie reviews have gradually become an important evaluation indicator for consumers to decide which movie to watch,make investment decisions,and other activities.However,under the temptation of huge interests,some investors,irrational fans and other stakeholders begin to fabricate a large number of spam movie reviews by themselves or the people they hired to improve the reputation of the target movies,or deliberately smear their competitive movies of the same period.If these movie review spam and spammer cannot be detected and handled timely,they will seriously destroy the normal atmosphere of movie reviews,mislead the choices of consumers,and bias the attention of the movie producer.Finally,the trust in movie reviews of all parties will lost ultimately and the normal film industry environment will be influenced.Therefore,conducting research on detecting movie review spam and spammer has important theoretical value and practical significance.Review spam detection is a hot research topic in academia and industry.Most of the existing research only focus on the review for products on e-commerce platforms.However,the research on movie reviews is little.There are some differences between movie review activities and product review activities: In the review activities for e-commerce products,spammers often appear during several periods of time,while non-spammers are relatively dispersed.But compared with e-commerce products,the selling cycle of movie tickets is very short,so the peak for the number of movie reviews written by spammer and non-spammer will appear in the same time period? movie reviews are more subjective compared to the reviews for e-commerce products.The existence of these differences poses challenges to the movie review spam and spammer detection task.Based on the existing research,this paper fully analyzes the characteristics of movie review activities,proposes labeling strategies for movie review spam and spammer,and labels corresponding datasets.Finally,this paper proposes corresponding detection models for the tasks of movie review spam and spammer detection,which use the graph convolutional neural network to improve the performance of the detection models.The research work of this paper is mainly divided into the following stages:(1)This paper proposes strategies to label movie review spam and spammer and completes the construction of the datasets.At present,there is little research focusing on movie review spam and spammer detection,and there are few labeling strategies and datasets for movie review spam and spammer detection.Therefore,this paper analyzes the differences between spam movie review detection tasks and spam product review detection tasks,proposes strategies to label movie review spam and spammer based on the characteristics of collusive attack behavior within users and abnormal review behavior,and constructs corresponding datasets.(2)This paper conducts research on movie review spam detection task.At present,there are still some shortcomings in the existing detection methods for movie review spam detection tasks,such as insufficient feature engineering and lack of considering the collusion attack behavior within users.Therefore,based on the characteristics of movie review activities,this paper revises and designs corresponding feature engineering to process the spam movie review datasets.Meanwhile,this paper utilizes GAT to encode the feature of interaction between users,proposes a movie review spam detection model based on feature engineering and GAT.This paper effectively detects movie review spam,and verifies the performance of the movie review spam detection model through experiments.(3)This paper conducts research on movie review spammer detection task.At present,there are still some shortcomings in the existing detection methods for movie review spammer detection tasks,such as not considering the information in heterogeneous graph for movie reviews.Therefore,based on the characteristics of movie review activities,this paper designs a heterogeneous graph for movie review to detect movie review spammer.Then,this paper uses the heterogeneous graph convolutional neural network and multi-layer perceptron to encode the explicit relationship due to common review activity and implicit relationship due to similar review activity and proposes a movie review spammer detection model based on HGCN.This paper effectively detects movie review spammer,and designs multiple experiments to verify the detection performance of the movie review spammer detection model proposed in this paper.