摘要
[研究目的]通过从评论文本和评论者两个维度对在线评论的特征进行挖掘,探究电商平台如何有效提高虚假评论识别的准确性,增加用户在线商品评论可信度,为消费决策提供参考。[研究方法]提出一种基于多维特征和SMOTE-RF模型的虚假评论识别方法。首先,通过识别商品虚假评论线索,构建多维特征向量,引入情感极性等特征并进行单调化处理;其次,在评分偏离度中添加了商品得分均值等统计指标来全面刻画虚假评论;最后,针对在线评论数据集中真实评论与虚假评论类不平衡问题,运用SMOTE算法优化随机森林分类模型,从而达到提高虚假评论识别准确性的目的。[研究结论]实验结果显示该方法在正负样本不平衡的虚假评论识别中具有更高的准确率、召回率及F1值。其中评分偏离度特征对虚假评论识别的影响最大,情感极性可作为识别的次要参考特征。因此,综合考虑在线评论多维特征和正负样本不平衡可帮助电商平台对虚假评论进行有效的过滤,为消费者提供更为可靠的评论数据。
[Research purpose] We explore how e-commerce platforms effectively improve the accuracy of fake review identification and increase the credibility of users' online product reviews by analyzing the characteristics of online reviews from the two dimensions of content and reviewers.The study provides reference for consumption decision-making.[Research method] We propose a method based on multi-dimensional features and SMOTE-RF model.Firstly,a multi-dimensional feature vector is constructed based on the characteristics of online fake comments.The emotion polarity feature is processed by monotonization.Secondly,a statistical index is introduced to the score deviation degree to describe the fake reviews.Finally,SMOTE is used to optimize the random-forest classification model to improve the imbalance of online review.It improves the accuracy of false comment identification.[Research conclusion] Several experiments are carried out based on the real review data of Jingdong Mall.Experimental results show that the proposed method has higher accuracy,recall and F-value.A comprehensive consideration of multi-dimensional features and the imbalance of samples can enable platforms to filter fake comments efficiently and help consumers to acquire more reliable review data.
作者
杜姗
杨敏
仇蓉蓉
Du Shan;Yang Min;Qiu Rongrong(Institute of Economics and Management,Xi'an University of Posts and Telecommunications,Xi'an 710061)
出处
《情报杂志》
CSSCI
北大核心
2023年第4期156-164,共9页
Journal of Intelligence
基金
陕西省自然科学研究计划项目“社交媒体用户的隐私保护研究”(编号:2022JQ-732)
陕西省教育厅智库项目“双循环格局下陕西数字经济生态系统演化路径研究”(编号:22JT040)的研究成果之一。
作者简介
杜姗,女,1986年生,博士,硕士生导师,研究方向:信息决策;杨敏,女,2001年生,研究方向:信息决策;仇蓉蓉,女,1986年生,博士,硕士生导师,研究方向:云计算、信息资源建设。
