A novel approach to stock price prediction: averaging open and close prices with LSTM

Stock price prediction is critical for investors and analysts because it helps them make more informed financial decisions and has a significant impact on investment strategies and market operations. Accurate prediction models are critical in the financial sector for navigating the complexities of market behaviour and maximizing returns. This research introduces a novel approach to stock price prediction by averaging the open and close prices, comparing its effectiveness to the traditional method of using closing prices alone. We employed a deep learning LSTM model for predicting stock prices of three major Korean companies: Samsung Electronics, Naver, and SK Hynix, using data collected from the KOSPI market over a period of 14 years from Yahoo Finance. Our methodology included creating a new feature an average of the open and close prices and evaluating its impact on prediction accuracy. Additionally, we used Optuna for hyperparameter optimization to enhance the model’s performance. The results demonstrate that our approach significantly enhances predictive performance. For example, in the case of Samsung Electronics, the normalized root means square error (RMSE) improved from 0.954 to 0.007, the Mean Absolute Percentage Error (MAPE) decreased from 2.25 to 1.75, and the R2 score increased from 0.949 to 0.962. Similar improvements were observed for Naver and SK Hynix. These findings suggest that averaging open and close prices provides a more accurate representation of daily stock price predictions.