摘要
网上购物已成为当前消费的重要途径,虚假评论的存在极大地降低了在线评论的可信度与参考价值。在对消费者权益带来严重损害的同时,也对网络购物平台秩序造成严重冲击。本文提出基于情感极性及多维特征的虚假评论识别方法,从虚假评论的情感特点出发,通过基于情感词典的方法判断在线评论的情感极性,实现虚假评论的初筛。定义关于虚假评论的10个特征变量,并使用优势比OR和逐步回归方法筛选8个最优变量子集进入模型的特征工程。最终使用Logistic回归实现虚假评论的识别,通过混淆矩阵检验,该识别方法取得了较好的效果。
Online shopping has become an important way of consumption,and the existence of false comments greatly reduces the credibility and reference value of online comments.While causing serious damage to consumer rights,it also has a serious impact on the order of online shopping plaforms.This article proposes a false comment recognition method based on emotional polarity and multidimensional features.Starting from the emotional characteristics of false comments,the emotional polarity of online comments is determined through an emotional dictionary based method,achieving the initial screening of false comments.Define 10 characteristic variables for false comments,and use odds ratio OR and stepwise regression methods to screen 8 optimal variable quantum sets into the feature engineering of the model.Finally,logistic regression was used to identify false comments,and through confusion matrix testing,the recognition method achieved good results.
作者
闵雪
MIN Xue(Zhejiang Vocational and Technical College of Commerce,Hangzhou 310053,Zhejiang)
出处
《江苏商论》
2024年第2期27-31,共5页
Jiangsu Commercial Forum
关键词
电子商务
虚假评论
情感极性
逻辑回归
e-commerce
False comments
Emotional polarity
logistic regression
