Research On Multi Dimensional Feature Fake Review Detection Algorithm Based On Deep Learning

With the rapid development of e-commerce,online shopping has gradually become the main way of people shopping.In the process of online shopping,the commodity comment information in the Electronic Business platform has become the main basis for the decision of the consumers ' online shopping.However,more and more fake reviews appear in front of consumers driven by profits,and because of the network water army and professional fake review writer,it is hard to distinguish fake review and real comments,so that consumers can not intuitively obtain buyers' purchase experience and feedback,information of service experience.With the continuous efforts of researchers,more and more detection methods were proposed.However,in the face of high simulation characteristics of today's fake review,the current methods have revealed its limitations:methods based on the content features using traditional machine learning depends highly on domain knowledge to extracting of language features in natural language processing,so that the algorithm has a poor ability of recognition and generalization,and the false judgment rate is higher.The artificial criticism has high simulation,such that it is difficult to distinguish the fake reviews only from comment contents.In view of the above problems,this paper proposes a multidimensional feature based fake review detection model using deep learning.This model introduces the bidirectional long time Memory network model(Bl-LSTM)in the deep learning to obtain the contextual semantic information features of the product comment contents instead of the artificial features selected by traditional machine learning.At the same time,the feature of the commentary content,such as the characteristic of the reviewer's behavior,the characteristic of the merchant's behavior and the Convolution Neural Network(CNN)are used to make the feature combination of the convolution kernel(filter)with different steps to help distinguish the fake reviews which are highly simulated.Finally,the attention mechanism(Attention model)is added to the proposed model based on deep learning to adjust the influence weights of multidimensional features to the final classification result,to improve the accuracy and recall of model detection.Finally the attention mechanism displays the high visual output of feature weights.so as to improve user confidence in the recognition of fake reviews.