Quantitative Study on Online Product Reviews Credibility

摘要：

　　In order to make the credibility of reviews on goods quantized, this paper, which is based on empirical study of factors on credibility, chooses four quantifiable feature dimensions(content integrity , emotion balance, timeliness and reviewer identity) by aiming at digital camera field to build credibility features library, develop software to quantify credibility features and adopt support vector regression (SVR) model to quantify review credibility.Then by grid search method, correlation parameter is adjusted and optimized model is obtained.The MSE is 0.005 832 2 and γ2is 0.461 089.The experimental results indicate when the difference value between predictor and actual value is in -0.1,0.1 , the forecast accuracy reaches 88.20%, it shows this model can realize the valid quantization of the high and low credibility on reviews.

关键词： products reviews credibility measure digital camera support vector machines machine-learning

作者: GONG Silan DING Shengchun

作者单位: School of Economics and Management, Nanjing University of Science &Technology, Nanjing 210094, China

会议类型: 国际会议

会议名称: 第一届信息获取与知识服务国际会议暨第六届搜索行为与用户认知研讨会

会议地点: 武汉

会议语种:英文

页码: 209-214

在线出版日期: 2014-10-10（万方平台首次上网日期，不代表论文的发表时间）