Implementing a novel strategy to mitigate the risk associated with stock market investments: An examination of the KOSPI instance

Abstract

A stock market is a market where people buy and sell shares of public companies. This index acts as a measure of the economic well-being of a country and shows the performance of companies and the general business environment. Stock prices are influenced by the relationship between supply and demand. Forecasting the stock market is a formidable undertaking due to the presence of unpredictable factors that impact price fluctuations. Rapid advances in AI and machine learning (ML) approaches, the availability of vast data, and the increasing processing power of machines provide opportunities to develop advanced methods for stock price prediction. The daily data of the Korea Stock Price Index was analyzed as part of this investigation. This study utilizes an enhanced Gated Recurrent Unit (GRU) in conjunction with Battle Royal Optimization (BRO) to forecast stock prices. Additionally, two other optimizers, namely Grey wolf optimization (GWO) and Moth flame optimization (MFO), have been used to optimize the hyperparameters of the model. Every model undergoes assessment based on certain criteria, revealing that the suggested model achieves the highest coefficient of determination value of 0.9963. This study used 5-fold cross-validation to BRO-GRU for stock price prediction to ensure generalizability across datasets. Wilcoxon-based statistical testing proved the model's superiority. The training time for GRU, GWO-GRU, MFO-GRU, and BRO-GRU methods was also measured. These findings show that the BRO-GRU model's predictive accuracy is higher than the compared methods and suggests competitive training efficiency and is better suited to accommodate the volatile nature of the stock market.