FINANCIAL MARKET FORECASTING USING NEURAL NETWORK AND IMMUNE APPROACHES
DOI:
https://doi.org/10.32782/2786-9024/v4i6(38).359123Keywords:
stocks, financial market, forecasting, multidimensional time series, transformer, dendritic artificial immune network, clustering, graph neural network.Abstract
Accurate stock price forecasting is a key task for investment decision support in volatile financial markets. Existing recurrent neural network approaches do not fully capture long-range dependencies and cross- market relationships, which reduces forecast quality on the volatile markets of 2022–2025 [1; 2]. This paper proposes a hybrid financial market forecasting model combining three components: a Temporal Fusion Transformer (TFT) for multivariate time-series encoding with interpretable attention; a Dendritic Artificial Immune Network (daiNet) for automatic stock clustering and adaptive relationship graph construction; and a Graph Neural Network (GNN) for joint learning of temporal and relational features. TFT, unlike LSTM, provides interpretable attention over different time horizons and explicitly models important market events. The model was validated on daily data of 16 NASDAQ technology companies over the 2022–2025 period, covering the 2022 tech crash and the 2023–2024 AI boom. Clustering identified three stable market clusters centered on eBay, Microsoft, and Amazon, reflecting distinct correlation patterns confirmed by heatmap analysis. Forecast quality was evaluated using mean squared error (MSE); the full (TFT + daiNet + GNN) configuration achieved an MSE of 1.41% on the test interval. The predicted returns were also used to generate an investment decision: for each day in the test set, the stock with the highest predicted return for the next period was selected. Experiments were conducted on daily OHLCV data for a set of liquid equities with a 1–5 day forecasting horizon and a 30-day TFT input window. Analysis of TFT attention weights revealed concentration on 5-day and 20-day horizons, corresponding to weekly and monthly trading cycles and providing actionable insights for practitioners. The absence of negative correlations across all 16 companies confirms broad market synchronization under shared macroeconomic shocks.
References
S. Agrawal, G. Das, A. Garg, “A Systematic Review on Graph Neural Network-based Methods for Stock Market Forecasting,” ACM Computing Surveys, vol. 57, no. 2, 2024. DOI: 10.1145/3696411.
T. Phaladisailoed, T. Numnonda, “Stock Price Prediction Using a Hybrid LSTM-GNN Model,” arXiv:2502.15813, 2025. DOI: 10.48550/arXiv.2502.15813.
R. Bhowmik, S. Wang, “Stock Market Volatility and Return Analysis: A Systematic Literature Review,” Entropy, vol. 22, no. 5, p. 522, 2020. DOI: 10.3390/e22050522.
D. Shah, H. Isah, F. Zulkernine, “Stock Market Analysis: A Review and Taxonomy of Prediction Techniques,” Int. J. Financial Stud., vol. 7 (2), 2019, 26. DOI: 10.3390/ ijfs7020026.
C. Krauss, X. A. Do, N. Huck, “Deep Neural Networks, Gradient-Boosted Trees, Random Forests: Statistical Arbitrage on the S&P 500,” European Journal of Operational Research, vol. 259, no. 2, pp. 689–702, 2017. DOI: 10.1016/j.ejor.2016.10.031.
H. Liu, S. Huang, P. Wang, Z. Li, “A review of data mining methods in financial markets,” Data Science in Finance and Economics, vol. 1, no. 4, 2021, pp. 362–392. DOI: 10.3934/DSFE.2021020.
K. Olorunnimbe, H. Viktor, “Deep learning in the stock market – a systematic survey of practice, backtesting, and applications,” Artificial Intelligence Review, vol. 56, 2023, pp. 2057–2109. DOI: 10.1007/s10462-022-10226-0.
M. M. Kumbure, C. Lohrmann, P. Luukka, J. Porras, “Machine learning techniques and data for stock market forecasting: A literature review,” Expert Systems with Applications, vol. 197, 2022, 116659. DOI: 10.1016/ j.eswa.2022.116659.
A. Singh, P. Gupta, N. Thakur, “An Empirical Research and Comprehensive Analysis of Stock Market Prediction using Machine Learning and Deep Learning Techniques,” IOP Conf. Series: MSE, 1022, 2021, 012098. DOI: 10.108 8/1757-899X/1022/1/012098.
Y. Guo, “Stock Price Prediction Using Machine Learning,” Södertörn University, Master Dissertation, 2022, 41 p.
Y. J. Chen et al., “A novel technical analysis-based method for stock market forecasting,” Soft Computing, vol. 22, 2018, pp. 1295–1312. DOI: 10.1007/s00500-016-2417-2.
B. Lim, S. Ö. Arık, N. Loeff, T. Pfister, “Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecasting,” Int. Journal of Forecasting, vol. 37 (4), 2021, pp. 1748–1764. DOI: 10.1016/j.ijforecast.2021.03.012.
A. Vaswani et al., “Attention Is All You Need,” in Advances in Neural Information Processing Systems 30 (NeurIPS 2017), pp. 5998–6008, 2017. DOI: 10.5555/3295222.3295349.
C. Zhao et al., “Stock Market Analysis Using Time Series Relational Models for Stock Price Prediction,” Mathematics, vol. 11 (5), 2023, 1130. DOI: 10.3390/math11051130.
H. Wang, Y. Zhang, J. Liang, L. Liu, “DAFA-BiLSTM: Deep Autoregression Feature Augmented Bidirectional LSTM network for time series prediction,” Neural Networks, vol. 157, 2022, pp. 240–256. DOI: 10.1016/ j.neunet.2022.10.009.
H. Widiputra, A. Mailangkay, E. Gautama, “Multivariate CNN-LSTM Model for Multiple Parallel Financial Time-Series Prediction,” Complexity, 2021, 9903518. DOI: 10.1155/2021/9903518.
M. Korablyov, S. Dykyi, O. Fomichov, I. Ivanisenko, D. Antonov, S. Lutskyy, “Hybrid Stock Analysis Model for Financial Market Forecasting,” in Proc. IEEE Int. Conf. on Computer Science and Information Technologies (CSIT), 2023, pp 1–4. https://doi.org/10.1109/ CSIT61576.2023.10324069.
