Stock price prediction using multiple valuation methods based on artificial neural networks for KOSDAQ IPO companies

Abstract

ABSTRACT It is difficult to predict future payoffs for initial public offerings (IPOs), since the multiple valuation method used to determine IPOs’ prices provides estimates by reflecting current sentiments in specific market environments. As our model reflects accounting information and stock price, we find that the mean absolute percentage error that verifies the accuracy of IPO stock valuation improves return on investment by 15% to 20%. This can help shareholders and investors accurately estimate stock prices and engage in efficient investment decision-making, while contributing to fintech by applying machine learning to traditional techniques to analyse investment opportunities and optimise trading strategies.