Stock Price Prediction Based on Phase Space Reconstruction and TCN

DOI: https://doi.org/10.62517/jse.202511413

Author(s)

Wenke Guo*, Keying Liu, Jingyun Zhao

Affiliation(s)

North China University of Water Resources and Electric Power, Zhengzhou, Henan, China *Corresponding Author

Abstract

To address the issue of stock prices exhibiting complex nonlinear characteristics due to the combined influence of multiple factors, which makes accurate prediction difficult using traditional statistical methods, this paper proposes a predictive model integrating Phase Space Reconstruction (PSR), an attention mechanism, and a Temporal Convolutional Network (PSR-ATT-TCN). First, multidimensional variables undergo PSR processing to obtain input variables. Subsequently, an attention mechanism determines the weight coefficients of each input variable. The Temporal Convolutional Network (TCN) then captures long-term dependency information to establish the price prediction model. Experimental results demonstrate that the PSR-ATT-TCN model achieves smaller errors in MAE, MSE, and RMSE metrics, exhibiting enhanced prediction accuracy and generalization performance.

Keywords

Stock Price Prediction; Phase Space Reconstruction; Temporal Convolutional Network; Attention Mechanism

References

[1]Jose C. Principe J C, Kuo J M.Dynamic modelling of chaotic time series with neural networks. Advances in Neural Information Processing Systems, 1995, 7: 311-318. [2]Chelani A B. Predicting chaotic time series of PM10 concentration using artificial neural network. International Journal of Environmental Studies, 2005, 62(2): 1-7. [3]Li Z G, Li L L, Hao J, et al. Short-term wind power prediction based on phase space reconstruction theory. Applied Mechanics and Materials, 2013, 364: 112-117. [4]Chen J, Yang S, Zhang D, et al. A turning point prediction method of stock price based on RVFL-GMDH and chaotic time series analysis. Knowledge and Information Systems, 2021, 63(10): 2693-2718. [5]Wang He, Chen Huifeng, Xiong Min, et al. Short-term wind speed prediction modeling fusing CEEMDAN and ICS-LSTM. Journal of Electronic Measurement and Instrumentation, 2022, 36(4): 17-23. [6]Fu Wenlong, Zhang Xuanrui, Zhang Hairong, et al. Research on ultra-short-term wind speed prediction based on INGO-SWGMN hybrid model. Journal of Solar Energy, 2024, 45(5): 133-143. [7]Yan J, Mu L, Wang L, et al. Temporal convolutional networks for the advance prediction of ENSO. Scientific reports, 2020, 10(1): 8055. [8]Bednarski BP, Singh A D, Zhang W, et al. Temporal convolutional networks and data rebalancing forclinical length of stay and mortality prediction. Scientific Reports, 2022, 12(1): 21247. [9]Wei X, Xu Y. Research on carbon emission prediction and economic policy based on TCN-LSTM combined with attention mechanism. Frontiers in Ecology and Evolution, 2023, 11: 1270248. [10]Li W, Jiang X. Prediction of air pollutant concentrations based on TCN-BiLS-TM-DMAttention withSTL decomposition. Scientific Reports, 2023, 13(1): 4665. [11]Zhang Tingyi, Huang Liqin, Chen Xiangxiang. Modern Information Technology Prediction, 2024, 8(17): 68-72. DOI:10.19850/j.cnki.2096-4706.2024.17.013.