基于商品评论的群体用户情感趋势预测研究  

Research on the Prediction of Group User Sentiment Trend Based on Commodity Comment

作　　者：周俊鹏 高岭[1,2] 曹瑞 高全力 郑杰 王海 ZHOU Junpeng;GAO Ling;CAO Rui;GAO Quanli;ZHENG Jie;WANG Hai(School of Information Science and Technology,Northwest University,Xi′an 710027,China;School of Computer Science,Xi′an Polytechnic University,Xi′an 710048,China)

机构地区：[1]西北大学信息科学与技术学院,陕西西安710027 [2]西安工程大学计算机科学学院,陕西西安710048

出　　处：《郑州大学学报（理学版）》2019年第4期23-29,共7页Journal of Zhengzhou University:Natural Science Edition

基　　金：国家自然科学基金项目(61672426,61572401);国家重点研发计划项目(2017YFB1002500)

摘　　要：提出了一种基于商品评论的群体用户情感趋势预测方法.首先,提出了基于BosonNLP的情感特征词识别算法,对时间维度下的商品评论信息进行特征选取;其次,使用群体用户多维特征向量构造多层感知器(MLP)模型进行情感分析;最后,融合评论时间和用户情感倾向值构建群体用户时序情感倾向序列,并通过长短时记忆网络(LSTM)模型进行时序情感趋势预测.在大规模真实数据集上的实验结果表明,MLP模型具有较好的分类效果;相比于现有的自回归(AR)模型,LSTM模型的平均均方差降低了79.06%,能够取得更加精准的预测结果.The group user sentiment trend prediction method which was based on commodity comment was proposed.Firstly,the analysis method of emotional characteristic words based on BosonNLP was put forward,to select the commodity comment information in the time dimension.Secondly,MLP was constructed by using group user multi-dimensional feature vector to analyze user′s sentiment.Finally,the group user time sentiment tendency sequence was constructed by integrating the comment time and the user′s sentimental trendency value.The sequence sentimental trend prediction was performed by using the LSTM model.Experimental results of large-scale real datasets indicated that the MLP model had a good classification effect.Compared with the existing autoregressive(AR)model,the average MSE of the LSTM was reduced by 79.06%,which could achieve a more accurate prediction result.

关 键 词：群体用户 商品评论 情感分析 时间序列 趋势预测 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]