Research On Spam Review Detection Based On Integrated Multi-feature

With the advent of Web2.0 and the continuous development of Internet technology,online shopping behavior has gradually become a social trend.Every major e-commerce platform will generate a large number of reviews every day,which provide important reference value for consumers and gradually become the main basis for consumers' purchase decisions.However,some illegal businesses will hire reviewers to give false descriptions of the goods,deliberately tout their own products or slander the comp's products,so that consumers can receive false feedback on the quality of the goods,and mislead consumers to make wrong consuming behaviors,seriously effect the healthy development of e-commerce,destroy the competition rules.Due to the strong confusion of spam reviews,it is unrealistic to identify a large number of commodity reviews by means of artificial identification.Therefore,how to effectively detect and identify spam reviews has become one of the problems that need to be solved in the development of the e-commerce.In this paper,the advantages of traditional machine learning and deep learning methods are comprehensively considered to detect and identify spam reviews.Two aspects of work in this paper are as follows:(1)A CNN-GRU based attention mechanism spam review detection and recognition model is proposed.In order to effectively identify spam reviews and make use of the advantages of deep learning algorithm in text feature learning,a hybrid CNN and GRU network attention mechanism model,is designed.The model consists of three parts: CNN network layer of word level feature learning and GRU network layer of sentence level feature learning.The attention mechanism is used to integrate the features in GRU network layer,which can analyze the review semantics from the semantic level and identify the spam reviews.In order to verify the validity of the model,an experiment on yelp public dataset was executed,and analyze the recognition effect under different model settings.The experimental results of the proposed model show that the model can effectively classify and identify spam reviews.(2)A model of spam reviews detection and recognition based on multi feature fusion is constructed.First of all,the statistical analysis of the data set is carried out,and four characteristic including n-gram feature,vocabulary feature,psychological feature and user behavior feature are proposed.Then CNN is used to process the extracted features and obtain the discrete feature vector.After that,fuse CNN to learn the sentence representation of word vector,and use two-way GRU and attention mechanism to encode and weight the sentence representation vector,so as to extract the deep semantic features of reviews text.Finally,discrete features and semantic features are combined to identify and detect spam reviews.Experiments shows that this model can efficiently recognize spam reviews,which can give right instruction information for consumers.