Research On Anomaly Comment Recognition Method Based On Multi-dimensional Feature Detection

With the rapid development of e-commerce,online shopping has gradually become one of the widely accepted shopping methods.At the same time,more and more online shopping users like to share their opinions and feelings about the purchased products and store services on the e-commerce platform after shopping.A large number of theoretical and empirical studies have shown that online shopping reviews have an important impact on consumer purchasing decisions and the branding of e-commerce brands and influence in online shopping scenarios.As a result,driven by interests,some unscrupulous merchants have adopted false opinions by hiring false commentators to improve their own credibility or discredit their competitors' reputation,and ultimately misleading consumers' purchasing decisions,resulting in a large number of false comments on the Internet,not only affecting Consumers' normal purchase decisions have seriously affected the healthy development of e-commerce.This problem has aroused widespread concern from all walks of life and needs to be resolved.Based on the above background,abnormal comment detection has become one of the research hotspots in the field of e-commerce and artificial intelligence in recent years.Domestic and foreign scholars have carried out a lot of research from the aspects of behavioral motive,formation path and automatic recognition of false comments.Based on the study of existing research results,this paper starts with the abnormal online shopping comment recognition problem,and starts with the characteristics analysis of the online shopping user's comment behavior and comment content,and proposes an abnormal online shopping comment recognition method based on behavior and content feature fusion.The main research work and the research results obtained are as follows:(1)A method for character recognition of normal comment groups based on information entropy is proposed.The method first determines the comment behavior attribute and the comment user classification,and then constructs the comment date entropy model of different categories of users based on the comment behavior attribute.Finally,the comment date entropy model is used to extract the behavior characteristics of different categories of normal comment groups,and then the abnormal comment behavior detection.Provide model and feature data support.(2)A normal comment text feature recognition method based on the subject content deviation degree calculation is proposed.Firstly,the method constructs the feature text feature set,and then constructs the topic deviation degree calculation model based on the text feature.Finally,the topic deviation degree calculation model is used to extract the topic deviation degree feature of the normal comment text,and then provides model and feature data support for the abnormal comment content detection.(3)An abnormal comment detection method based on the combination of comment behavior and comment content feature is proposed.The method is divided into two stages.The first stage combines the characteristics of the Japanese commentary behavior and the features of the Japanese text,and then uses the SVM classifier trained by the fusion feature to coarsely grain the daily abnormal comment behavior existing in the comment corpus.Identification;the second stage uses the feature fusion method to multi-feature fusion of the comment text,and then uses the merged text feature to perform SVM classifier training.Finally,the classifier is used to match the daily abnormal comment behavior identified in the first stage.The set of comments is used as input to perform false comment recognition based on text content.The experimental results show that the proposed method has better recognition accuracy than the traditional single dimension recognition algorithm.