Abstract

Stock price prediction remains a challenging problem in financial forecasting due to the nonlinear, noisy, and highly dynamic nature of market behavior, influenced by a complex interplay of economic events, investor sentiment, macroeconomic policies, and hidden temporal patterns. Accurate prediction of stock prices can significantly benefit individual investors, portfolio managers, and financial institutions by enabling more informed decision-making and risk management. However, the stochastic and non-stationary characteristics of financial time series make reliable forecasting extremely difficult using conventional statistical methods. This paper presents a Long Short-Term Memory (LSTM) based multivariate forecasting framework for predicting the next day's opening and closing prices of four major U.S. equities The proposed framework does not solely rely on raw OHLCV (Open, High, Low, Close, Volume) data. It enriches the input feature space through systematic technical analysis, generating 10 additional indicators including Simple Moving Averages at 5, 10, and 20-day windows (MA5, MA10, MA20), Exponential Moving Average (EMA20), Relative Strength Index (RSI), Moving Average Convergence Divergence (MACD) with its signal line, Bollinger Bands (upper and lower), Daily Return, and Volatility. The framework is evaluated using three standard regression metrics: Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and the coefficient of determination (R²). Experimental results across all four stocks demonstrate that the combination of technical indicator feature engineering with LSTM sequence learning produces meaningful forecasting performance.

References

• S. Hochreiter and J. Schmidhuber, "Long short-term memory," Neural Comput., vol. 9, no. 8, pp. 1735-1780, 1997.
• T. Fischer and C. Krauss, "Deep learning with long short-term memory networks for financial market predictions," Eur. J. Oper. Res., vol. 270, no. 2, pp. 654-669, 2018.
• M. Agrawal, A. A. A. Khan, and P. K. Shukla, "Stock price prediction using technical indicators: A predictive model using optimal deep learning," Int. J. Recent Technol. Eng., vol. 8, no. 4, pp. 7821-7829, 2019.
• S. D. P. et al., "Stock market prediction system using LSTM with technical indicators," IEEE Access, 2023.
• Y. Wen, P. Lin, and X. Nie, "Research of stock price prediction based on PCA-LSTM model," IOP Conf. Ser., Mater. Sci. Eng., vol. 790, p. 012109, 2020.
• T. Kim and H. Y. Kim, "Forecasting stock prices with a feature fusion LSTM-CNN model using different representations of the same data," PLoS ONE, vol. 14, no. 2, e0212320, 2019.
• A. Selvin, R. Vinayakumar, E. A. Gopalakrishnan, V. K. Menon, and K. P. Soman, "Stock price prediction using LSTM, RNN and CNN-sliding window model," in Proc. ICACCI, 2017, pp. 1643-1647.
• J. M.-T. Wu et al., "A graph-based CNN-LSTM stock price prediction algorithm with leading indicators," Multimedia Systems, vol. 27, pp. 377-400, 2021.