Building Consumer Confidence Index Based On Online Social Media Big Data

Arrival of the age of big data challenges to the traditional psychological research, it's expected to expand the areas of psychological research and improve the scientific system of psychology by combining big data technology with psychological problems. With the rapid development of internet technology, social media represented by twitter and blogs entered a period of rapid development. Related scholars have been aware of the tremendous value of social media information, and had applied it to various fields. Among them, the micro blog has writing simple, made fast, and other features, more and more users choose micro blog to express their views and attitudes.Consumer confidence also belongs to the category of psychology, reflecting the consumers' perception of current or future economic conditions, is an effective index of the economic and consumer trends. As a good platform for big data mining, micro blog can provide rich resource for the consumer confidence research. This study took blog text as the original data, build consumer confidence index user dictionary, calculated the consumer confidence index through different experimental conditions of the target words' frequency analysis. Explored the better way among "exact match" and "fuzzy match", tested the necessity of degree words and negative words. The conclusions are as follows:(1)Build consumer confidence index user dictionary. This dictionary includes 100 words, covering economic situation, interest rate, spending, employment, income, the quality of life. To a certain extent, this dictionary could represent subjective experience and attitudes for our current and future economic conditions.(2)Explored the predictive validity of target words' frequency at different ways of matching and different weights settings. Matching is divided into "exact match" and " fuzzy match", weights setting is divided into "considering degree words" and " considering degree words and negative words". The best way is "fuzzy match target words and consider degree words".(3)Verified the effectiveness of calculating the consumer confidence index through the target words' frequency analysis. Calculated the next month' consumer confidence index by validation datasets based on the best way, the difference between estimated value and official value is within 10%, so verified the effectiveness of this method.