뉴스와 댓글 텍스트를 활용한 공모주 상장일 시초가 형성 예측 및 영향성 분석: XGBoost 기반

This study assumes that news and comments can affect price formation immediately after IPO. The empirical analysis was conducted on 337 publicly traded companies newly listed from 2018.1 to 2022.7 to see if the initial price affects the formation immediately after the IPO listing by using the comment data including the interest of individual investors before the IPO and the news data in which the comment is generated. In order to utilize news and comment data, KR-FinBERT model was applied to news and comments posted on Naver News from the date of submission of each company’s declaration to the date of listing, and it was classified into positive, negative, and neutral, and then used as an emotional scoring variable. In addition, we analyzed the macroeconomic situation and individual and market responses using various variables affecting IPO price formation to determine whether the initial price is 1.0 times and 1.5 times higher than the final IPO price on the day of IPO. As a result of the study, when the initial price is more than 1.0 times higher than the final offer price, the performance is better than when the emotional variable is not included, but when the initial price is more than 1.5 times, the performance is lower than when the emotional variable is not included. Through this, it was confirmed that the emotional variables are helpful in predicting the initial formation compared to the final public offering, but not in predicting the excess increase.