基于评论特征时间序列的产品垃圾评论检测模型
首发时间:2016-05-27
摘要:产品垃圾评论检测是社交媒体情感分析的一项重要任务。针对评分倾向性和评论有用投票数对评论可信度产生影响的问题,本文提出了一种基于评论特征时间序列的产品垃圾评论检测方法。该方法首先通过定义找出商店产品中的可疑评论;然后对商店产品的评论相关特征建立多维时间序列图,通过异常点检测算法识别出时间序列图中各个特征维度上的共同突发区间;最后,从共同突发区间内找出符合可疑垃圾评论定义的评论。实验结果表明,与基准方法相比,本方法在准确率和召回率上分别提高了6.21%和18.83%。
For information in English, please click here
Detection of Product Review Spam Based on Review Time Series Feature
Abstract:Product review detection is an important task in social media sentiment analysis. Considering the impact on review reliability by the review rating tendency and review helpful vote, this paper proposes a new method to detect review spam based on review feature time series. Firstly, the suspicious reviews is located in store product by the means of suspicious review definition. Secondly, a multidimensional time series graph of the store product is constructed via review related features, and the common abnormal interval is found out through outlier detection algorithm. Finally, the suspicious review in common abnormal interval is detected as review spam. The experiment results show that compared with the baseline method, the accuracy and recall of our method improves by 6.21% , 18.83% respectly.
Keywords: Text Mining Sentiment Analysis Review Spam Dection
论文图表:
引用
No.4693741115215514****
同行评议
共计0人参与
勘误表
基于评论特征时间序列的产品垃圾评论检测模型
评论
全部评论0/1000