STOCK MARKET PREDICTION USING MACHINE LEARNING

This paper presented a complete machine learning pipeline for stock price forecasting and trading signal generation, combining classical ensemble models (Random Forest, XGBoost) with a deep learning sequence model (LSTM) and a Logistic Regression classifier within a single, quality-gated system. Scale-invariant feature engineering, walk-forward cross-validation, and an explicit minimum-quality threshold collectively address common pitfalls in financial machine learning, including non-stationarity, data leakage, and the deployment of unreliable models.   End-to-end evaluation on two instruments — COST and SPY — showed the LSTM consistently outperforming both tree-based models on regression error metrics, while the directional classifier and backtest results diverged sharply between the two instruments: near-chance performance on the individual equity (COST) versus a meaningfully profitable backtest on the diversified index fund (SPY). This finding underscores that the reliability of machine-learning-driven trading signals is not uniform across instruments, even when the underlying model architecture and feature pipeline are held constant — reinforcing the value of the system's transparent, per-stock accuracy and backtest reporting over a one-size-fits-all reliability claim.   As with all approaches to stock price prediction, the system's predictive power remains bounded by the inherent unpredictability of financial markets; no model in this study, nor in the broader literature, eliminates this fundamental limitation. Future work includes extending the evaluation to a larger universe of stocks across multiple sectors, integrating sentiment analysis from news and social media as an additional model input, extending the architecture to Transformer-based sequence models, supporting brokerage API integration for direct trade execution, and accounting for transaction costs and slippage in the backtest methodology

[1] M. I. K. Jabed, "Stock Market Price Prediction using Machine Learning Techniques," American International Journal of Sciences and Engineering Research, vol. 7, no. 1, Feb. 2024, doi: 10.46545/aijser.v7i1.308.   [2] L. N. Mintarya, J. N. M. Halim, C. Angie, S. Achmad, and A. Kurniawan, "Machine Learning Approaches in Stock Market Prediction: A Systematic Literature Review," Procedia Computer Science, vol. 216, pp. 96–102, 2023.   [3] D. Liu, A. Chen, and J. Wu, "Research on Stock Price Prediction Method Based on Deep Learning," in Proc. 2nd Int. Conf. Information Technology and Computer Application (ITCA), 2020, pp. 69–73, doi: 10.1109/ITCA52113.2020.00022.   [4] G. Bathla, "Stock Price Prediction Using LSTM and SVR," in Proc. 6th Int. Conf. Parallel, Distributed and Grid Computing (PDGC), 2020, pp. 211–214, doi: 10.1109/PDGC50313.2020.9315800.   [5] M. Bansal, A. Goyal, and A. Choudhary, "Stock Market Prediction with High Accuracy Using Machine Learning Techniques," Procedia Computer Science, vol. 215, pp. 247–265, 2022.   [6] Md. M. Akhtar, A. S. Zamani, S. Khan, A. S. A. Shatat, S. Dilshad, and F. Samdani, "Stock Market Prediction Based on Statistical Data Using Machine Learning Algorithms," Journal of King Saud University – Science, vol. 34, no. 4, art. 101940, 2022.   [7] R. Najem, M. Fakhouri Amr, A. Bahnasse, and M. Talea, "Advancements in Artificial Intelligence and Machine Learning for Stock Market Prediction: A Comprehensive Analysis of Techniques and Case Studies," Procedia Computer Science, vol. 231, pp. 198–204, 2024.   [8] A. Subasi, F. Amir, K. Bagedo, A. Shams, and A. Sarirete, "Stock Market Prediction Using Machine Learning," Procedia Computer Science, vol. 194, pp. 173–179, 2021.   [9] H. Wu, "Comparison of Random Forest and LSTM in Stock Prediction," Advances in Economics, Management and Political Sciences, vol. 86, no. 1, pp. 28–34, May 2024.   [10] K. Kashif and R. Ślepaczuk, "LSTM-ARIMA as a Hybrid Approach in Algorithmic Investment Strategies," arXiv:2406.18206, Jun. 2024.