Research on Stock Forecasting Based on GPU and Complex-Valued Neural Network
Accurately and rapidly forecasting stock index is a difficult and hot research in the economic field. In this paper, complex-valued neural network (CVNN) model is proposed to predict stock price. In order to improve the time of training CVNN model, a parallel particle swarm optimization (PSO) based on graphics processing unit (GPU) is proposed to optimize the complex-valued parameters of CVNN model. Shanghai stock exchange composite index is selected to demonstrate the performance of CVNN model with parallel PSO algorithm. The experiment results reveal that our proposed method could improve stock index accurately and reduce training time sharply.
KeywordsStock forecasting Neural network Complex-valued Graphics processing unit
This work was supported by the Natural Science Foundation of China (No. 61702445), the PhD research startup foundation of Zaozhuang University (No.2014BS13), Zaozhuang University Foundation (No. 2015YY02), and Shandong Provincial Natural Science Foundation, China (No. ZR2015PF007).
- 3.Pang, X., Zhou, Y., Wang, P., Lin, W., Chang, V.: An innovative neural network approach for stock market prediction. J. Supercomput. 1, 1–21 (2018)Google Scholar
- 6.Mei, S., He, M., Shen, Z.: Optimizing hopfield neural network for spectral mixture unmixing on GPU platform. IEEE Geosci. Remote Sens. Lett. 11(4), 818–822 (2013)Google Scholar
- 10.Kennedy, J.: Particle swarm optimization. In: Sammut, C., Webb, G.I. (eds.) Encyclopedia of Machine Learning. Springer, Boston (2002)Google Scholar
- 11.Chen, Y.H., Abraham, A.: Hybrid learning methods forstock index modeling. In: Kamruzzaman, J., Begg, R.K., Sarker, R.A. (eds.) Artificial Neural Networks in Finance, Health and Manufacturing: Potential and Challenges, IdeaGroup Inc. Publishers, USA (2006)Google Scholar