Method Of Recognition And Searching Network Spammers For Cultural Products

With the rapid development of the cultural industry and the rise of social networks,online reviews have gradually become an important evaluation index of consumer choice.However,the fake comments behavior of spammers will mislead the users into consumption and disrupt the normal development of cultural industry,hinder the improvement of the quality of the cultural industry.The existing recognition models and methods mainly focus on the recognition of network spammers for material goods and there are two main deficiencies in solving the identification of network spammers for cultural products:1.The existing network spammers recognition feature model has the problem of weak expression ability for network spammers for cultural products.In addition,the single-classifier can not have a high accuracy in identifying network spammers.2.What makes the identification of the network spammers have the problem of timeliness is the huge scale and dynamic changes of network spammers' behavior.In order to solve these problems,this thesis proposes a multi-perspective network spammers recognition method for cultural products and a network searching method for network spammers for cultural products based on multi-relationship fusion algorithm.The two methods can improve the accuracy and efficiency of network spammers recognition for cultural products respectively.In order to solve the problem of insufficient recognition performance on network spammers for cultural products with the existing network spammer feature model,this paper establishes a representative feature expression model of network spammers for cultural products.This thesis analyzes the three characteristics of cultural products,such as rich semantics,strict timeliness and network interaction.Then,it puts forward the comment subject similarity,average usefulness,behavior relevance,interest relevance,average evaluation enthusiasm and comprehensive quality evaluation from three perspectives of content,behavior and attribute.The new features are combined with the existing features and the redundant features are removed by feature selection method to form a new expression model.In view of the problem of low accuracy in identifying network spammers with single-classifier,XGBoost(Extreme Gradient Boosting)ensemble learning algorithm is used for spammer recognition on the basis of multiple perspectives.Experiments prove that the proposed feature expression model has a good degree of discrimination,and the multi-perspective XGBoost ensemble learning algorithm has a high accuracy rate for the network spammers recognition of cultural products.In order to solve the challenge of massive and real-time-increase user-data and dynamic changes of network spammers' behavior on the time efficiency of identifying network spammers,a search algorithm of spammer network for cultural products is proposed.Using the strong and weak relationship in social networks for reference,this thesis proposes a multi-relationship fusion algorithm of searceing spammer network for cultural products based on the attention-based relationship,f(?)d-based relationship and the same comment object according to the strong and weak relationship between users.The spammer groups are mined from the spammer individuals.The comparative experiments show that this method can greatly improve the time efficiency of searching network spammers for cultural products on large-scale dataset.