Machine Learning-Based Data Mining Method for Sentiment Analysis of the Sewol Ferry Disaster's Effect on Social Stress

Abstract

The Sewol Ferry Disaster which took place in 16th of April, 2014, was a national level disaster in South Korea that caused severe social distress nation-wide. No research at the domestic level thus far has examined the influence of the disaster on social stress through a sentiment analysis of social media data. Data extracted from YouTube, Twitter, and Facebook were used in this study. The population was users who were randomly selected from the aforementioned social media platforms who had posted texts related to the disaster from April 2014 to March 2015. ANOVA was used for statistical comparison between negative, neutral, and positive sentiments under a 95% confidence level. For NLP-based data mining results, bar graph and word cloud analysis as well as analyses of phrases, entities, and queries were implemented. Research results showed a significantly negative sentiment on all social media platforms. This was mainly related to fundamental agents such as ex-president Park and her related political parties and politicians. YouTube, Twitter, and Facebook results showed negative sentiment in phrases (63.5, 69.4, and 58.9%, respectively), entity (81.1, 69.9, and 76.0%, respectively), and query topic (75.0, 85.4, and 75.0%, respectively). All results were statistically significant (p < 0.001). This research provides scientific evidence of the negative psychological impact of the disaster on the Korean population. This study is significant because it is the first research to conduct sentiment analysis of data extracted from the three largest existing social media platforms regarding the issue of the disaster.