基于互联网文本情感分析的消费情感指数构建

Consumer Sentiment Index Construction Based on Text Mining and Sentiment Analysis on Online News

消费者信心指数主要反映消费者对于经济发展、就业状况、物价水平、生活状况、购房和投资6方面的信心情况。从以上6个维度出发,基于百度搜索采集2011—2017年消费者信心相关新闻文本174 606条,利用情感分析、深度神经网络学习等方法得到新闻文本的情感倾向,并以此为基础提出并构建了消费情感指数,用于反映中国消费者的信心情况,并将结果与传统问卷调查得到的消费者信心指数进行对比,研究发现,消费情感指数与传统消费者信心指数走势具有较强相似性,相关性约为0.7,对消费短期趋势判断更灵敏,对传统指数及经济景气指数具有一定的先导作用,体现了消费情感指数的可靠性、时效性,从侧面反映了舆论对消费者信心的影响。

The consumer confidence index mainly reflects the consumer confidence in the six aspects of economic development,employment status,price level,living conditions,purchase and investment.Starting from the above six dimensions,the article establishes an index of consumer confidence based on 174606 online news from 2011 to 2017 from ‘baidu’.Through methods of text mining,sentiment analysis and deep learning,we classify each news to a category of positive,neutral or negative and use the numbers of positive and negative tags quarterly to establish consumer sentiment index.The results show that the index based on online news is highly correlated with Chinese consumer confidence index derived from traditional way （CCI） with a correlation of 0.7.Besides,a trend of leading indicators on several points shows that the index is reliable as well as timely,the leading trend also shows that media and public opinion do have an effect on consumer confidence.