摘要
建立一个评论有用性模型,该模型能够对在线商品评论进行有用性预测.基于精心建立的情感词典,联合基于神经网络构建的商品属性词典.设计了合理的匹配算法,采用随机森林算法和五折交叉验证对评论有用性不同特征进行准确率、召回值和F指标的预测.结果显示评论有用性方差值能够很好地反应评论的有用性.评论有用性模型能够对评论进行准确的预测,所预测的结果可以为消费者提供有效的参考.
Build a model that can predict the usefulness prediction of online commodity reviews.Based on a carefully constructed emotional dictionary,unite the commodity attribute dictionary which based on neural network.It designs a reasonable matching algorithm,uses random forest algorithm and half off cross validation method to forecast the accuracy,recall value and F index of comment usefulness on the different characteristics.The results show that the variance yields of the comment usefulness can reflect the usefulness of reviews.The comment usefulness model is able to predict the reviews accurately,and the predicted results can provide an effective reference for consumers.
作者
王晗
李存林
杨世瀚
WANG Han;LI Cun-lin;YANG Shi-han(College of Software and Information Security;Guangxi Key Laboratory of Hybrid Computation and IC Design Analysis , Guangxi University for Nationalities, Nanning 530006 ,China)
出处
《广西民族大学学报（自然科学版）》
CAS
2018年第1期70-75,94,共7页
Journal of Guangxi Minzu University ：Natural Science Edition
基金
国家自然科学基金(11371003
11461006)
广西科技基地和人才专项(2016AD05050
广西自然科学基金面上项目(2014GXNSFAA118359)
广西"八桂学者"专项资助
广西民族大学研究生教育创新计划资助(gxunchxps201676)
