빅데이터에 나타난 감성 분석

초록

We study thoroughly by looking into nine different sentiments found in approximately 190 million pieces of Big-data gained from January 1st, 2011 to January 4th, 2013. In the past, it was not easy to extract the sentiments and because of that, until now, any influences that the sentiments could actually have on the stock market have been neglecting. In the study, with the sentiment references provided by Daum-soft, features of the sentiments were examined by autocorrelation analysis, principle component analysis and VAR. According to the results, we find that the sentiments are observed to have some regular patterns. In other words, the findings from the autocorrelation analysis prove autocorrelation and period of the sentiments while the results from the principle component analysis report that the nine sentiments could be connected with positivity and negativity. Lastly, via VAR, the sentiments appeare to have negative autoregressive parameters as they would be affected by each other at various lag-times. Those results from the analyses indicate that the sentiments with stock information appearing on Big-data would integrate with changes in the stock market as they can be possibly estimated based on values from the past.

키워드