Consumer Confidence Index Prediction Based On Web Search Data

Consumer confidence index(CCI) is a comprehensive reflection of consumer and quantify the evaluation of the current economic situation and the economic outlook, income level,income expectations and consumer psychological state of subjective feelings.And CCI is an important basis for monitor changes in the economic cycle, which has aroused wide public attention. The CCI survey methodology is still sampling method, the presence of the sample coverage is too small, lacking of objectivity, published data is not timely and other defects.And now the CCI evaluation system exist many differences, so as to promote in-depth impact consumer confidence. Thus we need to raise more new survey method. The popularity of the internet has provided us with a massive data base, and how to use these data for assisting decision-making needs of researchers becomes a hot research issue.This paper from the perspective of consumer buying behavior starting to search the data relations between network and CCI. Combined with the principle of the preparation of the National Bureau of CCI divided it into six indicators, obtain word search data associated with the use of key indicators from January 2014 to December 2015 by Baidu search engine. And use statistical methods such as principal component analysis, regression, BP neural networks for searching data and CCI made an empirical analysis, mainly concluded as follow:(1) The R2 of predict model of CCI build web-based keyword search data reached 0.828,and there exist of long-term and stable co-integration relationship between Web search keyword data and CCI. The words of related searches increase will cause a rise in the CCI.(2) Using a network search data can predict CCI earlier than the traditional way of questionnaire about a month. It has a strong timeliness, and can save a lot of labor costs.