USING AN EVOLVING GRAPH NEURAL NETWORK FOR PRICE FORECASTING ON THE STOCK MARKET
DOI:
https://doi.org/10.35546/kntu2078-4481.2025.2.2.24Keywords:
forecasting, financial market, stocks, relationship, connection graph, evolution, graph neural network, artificial immune networkAbstract
Stock price forecasting is essential for making informed investment decisions in the financial market. Analyzing financial market movements and stock price behavior is extremely complex due to the dynamic, nonlinear, non-stationary, non-parametric, and chaotic nature of markets. Various approaches are used to analyze stocks for the purpose of financial market forecasting. Traditional methods based on time series information for stocks of one company do not take into account the relationships between stocks of other companies. The use of graph neural networks (GNN) for this purpose, in which time series relationships are represented as a graph structure and variables are represented as graph nodes, significantly improves the accuracy of forecasting. Existing forecasting methods usually assume that the structure of the relationship graph, which is described by the relationship matrix and determines the aggregation method of the graph neural network, is fixed by definition. Therefore, they cannot effectively account for dynamic changes in the relationship graphs. In this paper, the use of an evolving graph neural network is proposed for forecasting stock prices in the financial market.To obtain dynamic correlations between price movements in financial time series, a cluster-based relationship graph is constructed, the generation and evolution of the structure and parameters of which are implemented using a dendritic artificial immune network (DaiNet). For each generated cluster of the relationship graph, price coding is performed using transformers to determine price information. Then, messages from the relationship graph structure and input time sequences are aggregated based on the use of the attention layer of the time graph. The final forecast of the future price movement of each stock is performed at the last layer of the GNN using a multilayer perceptron. Experimental studies have shown that the use of dynamic graph construction and stock clustering based on DaiNet tracks the temporal nature of the relationships between stocks and improves forecasting performance.
References
Shih S.-Y., Sun F.-K., and Lee H.-Y. Temporal pattern attention for multivariate time series forecasting. Machine Learning 108, 8, 2019, pp. 1421–1441. DOI: 10.48550/arXiv.1809.04206
Patel M., Jariwala K., and Chattopadhyay C. A Systematic Review on GNN-based Methods for Stock Market Forecasting. ACM Computing Surveys, Vol. 57, Iss. 2, Art. No 34, 2024, pp. 1–38. https://doi.org/10.1145/3696411
Cheng D., Yang F., Xiang S., and Liu J. Financial time series forecasting with multi-modality graph neural network. Pattern Recognition 121, 2022, 108218. https://doi.org/10.1016/j.patcog.2021.108218
Li W., Bao R., Harimoto K., Chen D., Xu J., and Su Q. Modeling the Stock Relation with Graph Network for Overnight Stock Movement Prediction. In IJCAI, 2020, pp. 4541–4547. https://doi.org/10.24963/ijcai.2020/626
Jin Z., Yang Y., & Liu Y. Stock closing price prediction based on sentiment analysis and LSTM. Neural Computing and Applications, 32 (13), 2020, pp. 9713–9729. https://doi.org/10.1007/s00521-019-04504-2
Liu W., Zhang Y., Wang J., He Y., Caverlee J., Chan P., Yeung D.S., and Heng P.-A. Item Relationship Graph Neural Networks for E-Commerce. IEEE Transactions on NNLS, 2021, pp. 1–15. DOI: 10.1109/TNNLS.2021.3060872
Wu Z., Pan S., Chen F., Long G., Zhang C., and Philip S.Y. A comprehensive survey on GNN. IEEE Transactions on NNLS, 32, 1, 2020, pp. 4–24. https://doi.org/10.48550/arXiv.1901.00596
Sawhney R., Agarwal S., Wadhwa A., and Shah R. Deep Attentive Learning for Stock Movement Prediction from Social Media Text and Company Correlations. In Proceedings of the Conference on Empirical Methods in NLP, 2020, pp. 8415–8426. DOI: 10.18653/v1/2020.emnlp-main.676
Wu Z., Pan S., Long G., Jang J., and Zhang C. Graph wavenet for deep spatial-temporal graph modeling. arXiv preprint arXiv:1906.00121, 2019. https://doi.org/10.48550/arXiv.1906.00121
Zhang Q., Chang J., Meng G., Xiang S., and Pan C. Spatio-temporal graph structure learning for traffic forecasting. In Proceedings of the AAAI Conference on Artificial Intelligence, V. 34, 2020, pp. 1177–1185. DOI: 10.1609/aaai.v34i01.5470
Korablyov M., Fomichov O., Antonov D., Dykyi S., Ivanisenko I., and Lutskyy S. A hybrid stock analysis model for financial market forecasting. IEEE Proceedings of the 18th Intern. Confer. on Computer Science and Information Technologies, 2023, pp. 1–4. DOI: 10.1109/CSIT61576.2023.10324069
Xu M., Dai W., Liu C., Gao X., Lin W., Qi G. J., and Xiong H. Spatial-temporal transformer networks for traffic flow forecasting. ArXiv abs/2001.02908, 2020. https://doi.org/10.48550/arXiv.2001.02908
Korablyov M., Fomichov O., Ushakov M., and Khudolei M. Dendritic Artificial Immune Network Model for Computing. Proceedings of the 7th International Conference on Computational Linguistics and Intelligent Systems (CoLInS 2023), 2023, pp. 206–217. https://ceur-ws.org/Vol-3403/
