머신러닝모형을 이용한 글로벌 자동차 기업의 주가 예측 비교

Abstract

As the stock price decoupling between domestic and foreign stock markets is prolonged, the so-called ‘Seohak ants’ are emerging to expand investing targets to the global stock market. This study is to predict the stock prices and volatility of global automotive companies and analyze whether economic value appears through performance analysis of investment strategy. The investment target was selected as automakers in Germany, the U.S., Japan, and Korea, which are advanced countries in the global automotive industry and fiercely competing as the paradigm of electric and self-driving cars emerged. Machine learning models were used to predict the stock prices, which show good results in predictive studies of time series data such as irregular and nonlinear moving stock prices. The empirical results from January 2011 to September 2022 data are as follows. First, daily low prices showed the highest importance score in feature variables for stock price prediction. Second, the random forest model in stock price prediction and the support vector regression model in volatility prediction showed excellent predictive power. Third, the investment strategy using the predicted stock price showed higher profitability than the benchmark, finding the economic value of the predicted stock price. This study is of academic and practical significance in that it predicted the stock prices volatility of the global automotive industry by using the machine learning models in the face of expanding global stock investment. In future studies, it is necessary to diversify the investment target and add sophisticated prediction techniques to study to improve the predictive performance.