摘要
论文以100个刷客(垃圾评论者)和100个正常评论者的历史评论数据作为研究对象,结合淘宝刷客的特点提取了14个用于刷客识别的特征,采用SVM算法和KNN算法构建分类模型并使用两种模型对淘宝网上的刷客进行识别。研究发现:两种分类模型对淘宝刷客识别的效果都很理想,其中SVM明显优于KNN,其分类模型对刷客识别的精确率达88%,召回率达100%。
This paper takes the historical reviews of 100 spammers and 100 normal buyers as the research subjects, and extracts 14 features according to Taobao spammers' characteristics. We use the SVM algorithm and KNN algorithm to respectively construct the classification model to i- dentify spammers. The result shows that the detection effect of both classification models are ver- y satisfactory; SVM is significantly better than KNN, its precision rate is 88% and recall rate is 100%.
出处
《科学决策》
CSSCI
2015年第9期79-94,共16页
Scientific Decision Making