| People are increasingly inclined to the way of online shopping and consumption,before purchasing products or services,people will usually look up the relevant reviews for reference.Undesirable shops or businessman to control public opinion to publish fake reviews to deliberately raise or slander,so that a large number of fake reviews mislead consumers to make risky purchase decisions,seriously disrupting the balance of the e-commence market.The existing fake review detection methods are facing with the severe challenge in the face of the increasingly sophisticated and difficult to judge,so it is very important to study the fake review.Existing fake review recognition approaches are facing the problems of not considering the effective review data set and calculating the score of fake probability doesn't reflect the importance of the features,so research fake review recognition based on graph model is proposed.Firstly,this paper uses the algorithm based on rating behavior to detect abnormal objects from the original review data,and then uses the algorithm based on feature weight calculation of Meta-path to identify the fake review about the abnormal objects.Aiming at the detection of the abnormal objects,a detection algorithm based on the difference of rating and kernel density estimation are designed respectively.And aiming at the recognition of fake reviews,a fake review recognition algorithm based on feature weight calculation of Meta-path(FRDA),constructs the review data set about abnormal objects into a review heterogeneous information network,uses the classification method of graph model to recognize fake reviews,and solves the problems of the overall sparseness due to fake reviews as well as the problem of unequal and characteristic weight calculation process.At last,the Yelp data set is experimented and tested.The FRDA algorithm is compared with SpEagle and ICF algorithms.The experiment shows that FRDA algorithm can effectively improve the accuracy of test results. |
PDF Full Text Request