Stock price prediction research has been actively underway since the past. Predicting the direction of the stock price movement is an attractive study that can be applied to actual investment. Even a little improvement in stock price performance can lead to large profits from actual investment. Nonetheless, predicting a random, volatile stock market poses many challenges. With the development of deep learning methodologies, there have been various attempts to apply them to predicting stock prices. This study aims to improve stock price direction prediction performance by using the Generative Adversarial Network (GAN) model, which is widely used in the image field, and the LSTM model, which is widely used for time series data prediction. The proposed LSTM-GAN model compared performance with that of benchmark models using the KOSPI 200 index, Korea’s representative stock index.