Research On False Comment Detection Method Combining The Content Of The Comment Text And The Behavior Characteristics Of The Commenter

With the development of the Internet,especially the emergence of e-commerce,people prefer online shopping,and it has become one of the most important ways for individuals.However,since it is difficult to have a true understanding of the purchased products in online shopping,consumers often use the product online review as the basis for purchasing the product.Driven by the interests,some unscrupulous merchants will use their professional writers to make fake praise reviews for their own goods or to give fake bad review to other.It is difficult for consumers to identify these spams.Therefore,in order to identify spam review effectively,some scholars use machine learning methods to identify and achieved some results.This paper studies spam detection further,combines the content of the review text with the behavior characteristics of the reviewers,and proposes a spam review recognition method based on multi-dimensional weight selection and a spam review recognition method based on double convolutional neural network,and then effectively merged two methods-a spam review recognition method based on multi-weighted double convolutional neural network.The multi-feature weights method for spam detection,mainly by extracting the feature content of the review text and the behavior characteristics of the reviewer,the random forest is used to find the weight value of each feature,and the feature is multiplied by its corresponding weight,and then the weight will be weighted.The feature uses a classification algorithm to identify spam reviews.The spam detecting method based on double convolutional neural network performs different processing on the comment text and the reviewer’s behavior characteristics.On the one hand,the content of the review text is expressed in the form of a word vector,and the feature is put into a convolutional neural network to transform it into a sentence vector.On the other hand,reviewer behavioral characteristics are extracted,and convolutional neural networks are used to combine and filter reviewer behavioral characteristics.Then,two parts extracted features are spliced,and finally the spliced features are classified using a classification algorithm.The spam detecting method based on double convolutional neural network with multi-feature weights is a fusion of the above two methods.On the one hand,the expression of the word vector is weighted,and the part of speech and the word frequency-inverse document frequency(TF-IDF)feature are added to the word vector representation.On the other hand,random forests are used to determine the weight of the reviewer’s behavioral features and weighted.The above-mentioned weighted features are processed using a method of double convolutional neural network.This paper has the research results of spam detection,which proves the effectiveness of the proposed algorithm and provides new research ideas and effective research identification methods for spam detection research.