기계학습기법을 이용한 코스닥 신규 공모주 상장일 시초가 예측에 관한 연구

Abstract

The purpose of this study is to predict the initial price of new public offering stocks on the listing date for short-term excess returns of public offering stocks due to the IPO underpricing factor. This study seeks to extract the characteristic factors related to the determinants of the public offering price and derive the factors that affect the public offering price. Specifically, for IPOs to be listed on the KOSDAQ market, the public offering price decision influencing factors are derived and artificial intelligence techniques (neural networks, decision trees) are used to build a public offering price prediction model. Research variables related to the public offering price prediction model include sales, net profit, largest shareholder’s share, capital stock, number of listed stocks, sales ratio of old stocks, lead company size, institutional subscription competition rate, general subscription competition rate, mandatory holding commitment rate, distribution Available volume, public offering amount, public offering price, public offering price discount rate, etc. As a result of the study, the main variables to be considered for the prediction of the public offering price were the general subscription competition rate and the public offering price. Variables included in the ranking of importance, such as capital stock, net income, commitment ratio of obligatory holdings, and the largest shareholder's equity ratio, were also extracted as factors to be considered.