학술논문
트럼프 취임 관련 국내 언론에서 나타난 감성과 거시 경제 지수 간 영향 관계 : 텍스트 마이닝을 적용하여
An Analysis of the Influences between Sentiment Values of Korean Online News and Macroeconomic Indicators Using Text Mining
An Analysis of the Influences between Sentiment Values of Korean Online News and Macroeconomic Indicators Using Text Mining
Document Type
Article
Author
Source
언론과학연구 / Journal of Communication Science. Sep 30, 2018 18(3):129
Subject
Language
Korean
ISSN
1598-2653
Abstract
본 연구는 감성 분석과 더불어 그랜저 인과 검정 및 벡터자기회귀모형 분석을 통해 어떠한 거시 경제 지수가 일정기간의 시차를 두고 뉴스 본문의 감성 값에 영향력을 끼치거나 영향을 받는지 확인하였다. 연구 대상이 되는 데이터는 2016년 1월 1일부터 2017년 6월 30일까지 보도된 것으로, 여러 언론사 뉴스를 제공해주는 네이버 뉴스에서 수집하였다. 연구 결과 한국의 언론에서 구한 감성 지수와 선후행 관계가 있는 경제 지표는 금 현물 지수, 코스닥 지수(KOSDAQ), 코스피 고가 지수(KOSPI high index), KRX 100으로 나타났다. 본 연구는 실시간으로 배포되는 온라인 뉴스 기사에서 가장 최근에 실시된 미국 대통령 선거를 주제로 진행한 연구라는 점에서 시의성이 크다. 그리고 감성 분석에 그치지 않고 거시 경제 지수들과의 상관관계를 탐구하였다는 데에 의의가 있다.
It was surprising that Donald Trump won the 58th American presidential election against the predictions of most press and opinion polling agencies. No one could expected the fluctuation in global economic indicators after the presidential election, so people have begun to focus on the reasons for Trump’s triumph. However, so far most analysts have only offered political or economic opinions and discourse. It is hard to properly capture economic changes in terms of Trump’s presidential election pledges such as “America First Policy”, “Trumpnomics”, and “Trump care”. Sentiment analysis conducted on data obtained using text mining is a useful way to analyze vast amounts of information, so it has been conducted on many different types of text, from product reviews to movie reviews. Studies related to sentiment analysis, Granger causality test and vector autoregressive model analysis simultaneously have not been conducted in South Korea. Studies that only focus on the relationship between economic indicators and sentiment indices of news pieces cannot determine why stock prices change in various contexts. In South Korea, not much research has been conducted on the correlation between Korean economic indicators and sentiment indices of South Korean internet text content. Therefore, it is necessary to study the connection between South Korean economic indicators and South Korean online news content In this study, sentiment analysis was conducted on news articles to generate sentiment indices on which a Granger causality test and vector autoregressive model analysis were conducted to determine the degree to which the sentiment values of online news text content can be used to predict changes in macroeconomic indicators influence to understand the background of the most recent U.S. presidential election more clearly. First, sentiment analysis was conducted on various news articles from Naver News for Korean-language news from January 1, 2016 to June 30, 2017. A Granger causality test was then run on the sentiment value data and macroeconomic indicators to determine the relationship between them. Sentiment analysis showed that Korean-language news had a vacillating value starting in March 2016 until it became decisively negative in early September 2016. In early November 2016, when the U.S. presidential election was held, the slope suddenly increased only to turn negative in mid-April 2017. Finally, the gold spot, KOSDAQ, KOSPI high, and KRX 100 in-dices were revealed as representative South Korean economic in-dicators for relating with sentiment values. These results sug-gested that the sentiment of South Korean news related to the U.S. presidential election was preceded or followed by changes in South Korea’s representative integrated stock price and gold spot indices. The main contribution of this study is that its discovery of why sentiment values can be used to make predictions about changes in economic when it comes to analyzing relation between senti-ment values and economic indicators.
It was surprising that Donald Trump won the 58th American presidential election against the predictions of most press and opinion polling agencies. No one could expected the fluctuation in global economic indicators after the presidential election, so people have begun to focus on the reasons for Trump’s triumph. However, so far most analysts have only offered political or economic opinions and discourse. It is hard to properly capture economic changes in terms of Trump’s presidential election pledges such as “America First Policy”, “Trumpnomics”, and “Trump care”. Sentiment analysis conducted on data obtained using text mining is a useful way to analyze vast amounts of information, so it has been conducted on many different types of text, from product reviews to movie reviews. Studies related to sentiment analysis, Granger causality test and vector autoregressive model analysis simultaneously have not been conducted in South Korea. Studies that only focus on the relationship between economic indicators and sentiment indices of news pieces cannot determine why stock prices change in various contexts. In South Korea, not much research has been conducted on the correlation between Korean economic indicators and sentiment indices of South Korean internet text content. Therefore, it is necessary to study the connection between South Korean economic indicators and South Korean online news content In this study, sentiment analysis was conducted on news articles to generate sentiment indices on which a Granger causality test and vector autoregressive model analysis were conducted to determine the degree to which the sentiment values of online news text content can be used to predict changes in macroeconomic indicators influence to understand the background of the most recent U.S. presidential election more clearly. First, sentiment analysis was conducted on various news articles from Naver News for Korean-language news from January 1, 2016 to June 30, 2017. A Granger causality test was then run on the sentiment value data and macroeconomic indicators to determine the relationship between them. Sentiment analysis showed that Korean-language news had a vacillating value starting in March 2016 until it became decisively negative in early September 2016. In early November 2016, when the U.S. presidential election was held, the slope suddenly increased only to turn negative in mid-April 2017. Finally, the gold spot, KOSDAQ, KOSPI high, and KRX 100 in-dices were revealed as representative South Korean economic in-dicators for relating with sentiment values. These results sug-gested that the sentiment of South Korean news related to the U.S. presidential election was preceded or followed by changes in South Korea’s representative integrated stock price and gold spot indices. The main contribution of this study is that its discovery of why sentiment values can be used to make predictions about changes in economic when it comes to analyzing relation between senti-ment values and economic indicators.