Author(s)

MingXi Ma

Affiliation(s)

School of Statistics and Data Science, Shanghai University of International Business and Economics, Songjiang 201600, Shanghai, China.

Corresponding Author

MingXi Ma

ABSTRACT

Predicting stock prices with high accuracy continues to be a major challenge in financial markets, primarily because of the intricate, non-linear, and highly volatile characteristics of price movements. Traditional statistical methods and standard long short-term memory (LSTM) networks exhibit limitations in capturing temporal dependencies and identifying critical features that significantly influence price movements. To address these challenges, this paper proposes a novel bidirectional LSTM with attention mechanism (BiLSTM-Attention) model for stock price prediction. The proposed model employs bidirectional LSTM layers to process time-series data in both forward and reverse directions concurrently, thereby capturing a more complete picture of past and potential future trends. Additionally, a self-attention mechanism is incorporated to dynamically allocate weights across time steps, enabling the model to focus on salient features that exert substantial influence on price fluctuations. Experimental validation is conducted using real-world stock price data from American International Group (AIG). Results demonstrate that the proposed BiLSTM-Attention model significantly outperforms baseline models across all evaluation metrics, validating the effectiveness of combining bidirectional processing with attention mechanisms for stock price forecasting. The proposed approach offers a stable and effective method for predicting stock prices in the short term.

KEYWORDS

Stock price prediction; Bidirectional LSTM; Attention mechanism; Deep learning; Time series forecasting

CITE THIS PAPER

MingXi Ma. Time-series forecasting of stock price via bidirectional LSTM-attention neural architecture. World Journal of Information Technology. 2025, 3(6): 36-42. DOI: https://doi.org/10.61784/wjit3074.

REFERENCES

[1] Henrique M B, Sobreiro A V, Kimura H. Literature Review: Machine Learning Techniques Applied to Financial Market Prediction. Expert Systems with Applications, 2019, 124: 226-251.

[2] Zhao Tingting, Han Yajie, Yang Mengnan, et al. A Review of Machine Learning-Based Methods for Time Series Data Forecasting. Journal of Tianjin University of Science and Technology, 2021, 36(5): 1-9.

[3] Mailagaha M K, Christoph L, Pasi L, et al. Machine Learning Techniques and Data for Stock Market Forecasting: A Literature Review. Expert Systems with Applications, 2022, 197.

[4] Xing Weichen. Stock Forecasting in the Big Data Era. China Business Review, 2020(3): 31-32.

[5] Vuong H P, Phu H L, Nguyen V H T, et al. A Bibliometric Literature Review of Stock Price Forecasting: From Statistical Model to Deep Learning Approach. Science Progress, 2024, 107(1).

[6] Bambang S, Derwin S. Utilizing BERT and CNN-LSTM in Stock Price Prediction Using Data Sentiment Analysis and Technical Analysis of Stock and Commodity. Bulletin on Innovative Computing, Information, and Control, 2023, 17(2): 171-176.

[7] Wang Haiyao, Wang Jianxuan, Chen Lihui, et al. A Stock Closing Price Prediction Model Based on CNN-BiSLSTM. Complexity, 2021.

[8] Yadav H, Thorat A. NOA-LSTM: An Efficient LSTM Cell Architecture for Time Series Forecasting. Expert Systems with Applications, 2024, 238.

[9] Yang Zheng, Wu Haocheng, Zhang Jing, et al. Can Dimension Reduction Data ForecastingEnhance the Predictive Performance of Stock Excess Returns?. Journal of Econometrics, 2023, 3(3): 828-847.

[10] Mao Yueyue, Zhang Qiuyue. Research on Machine Learning-Based Stock Forecasting Methods. Modern Computer, 2020(23): 44-47.