Academic Journal of Computing & Information Science, 2022, 5(10); doi: 10.25236/AJCIS.2022.051007.

An Improved ARIMA Method Based on Hybrid Dimension Reduction and BP Neural Network

Author(s)

Longhui Mo

Corresponding Author:
Longhui Mo
Affiliation(s)

Mathematics and Physics College, Chengdu University of Technology, Chengdu, 610059, China

Abstract

In order to solve the problem that the ARIMA model cannot well fit the prediction of time series with high dimension and high noise, this paper proposes a method based on the combination of hybrid reduction and BP neural network. Taking the stock price as an example, the proposed method takes the intraday price as auxiliary information and uses PCA and KPCA to extract linear and nonlinear features of it respectively, and the dimensionally reduced features are then used as the input variable. BP neural network model was used to fit the residual error between the real value and the predicted value of the ARIMA model. Finally, add up the closing price prediction value obtained by the ARIMA model and the residual error prediction value obtained by BP neural network model for the final closing price prediction value. The empirical results show that compared with the ARIMA model, the proposed method has better prediction performance and fitting accuracy, and has certain robustness. This method can also be extended to other practical problems such as average temperature prediction and port ship flow prediction.

Keywords

ARIMA; stock price prediction; dimension reduction; BP neural network

Cite This Paper

Longhui Mo. An Improved ARIMA Method Based on Hybrid Dimension Reduction and BP Neural Network. Academic Journal of Computing & Information Science (2022), Vol. 5, Issue 10: 41-47. https://doi.org/10.25236/AJCIS.2022.051007.

References

[1] Yingchao Zhang, Yingjun Sun. Empirical research on Shanghai Stock Index Analysis and Prediction based on ARIMA Model [J]. Economic Research Guide, 2019(11):5.

[2] Hong Cai, Rongyao Chen. Research on Stock Price Prediction based on PCA-BP Neural Network [J]. Computer Simulation, 2011, 28(3):4.

[3] Song Li, Lijun Liu, Man Zhai. Optimization of BP Neural Network for Short-term Traffic Flow Prediction by Improved Particle Swarm Optimization [J]. Systems Engineering-Theory & Practice, 2012, 32(9):2045-2049.

[4] Jianfeng Guo, Yu Li, Dong AN. Short-term Stock price Prediction based on LM Genetic Neural Network [J].Computer Technology and Development, 2017, 027(001):152-155,159.

[5] Joseph A, Larrain M, Turner C. Daily Stock Returns Features and Forecastability [J]. Procedia Computer Science, 2017, 114:481-490.

[6] Fengjuan, Miao, Tongri, et al. Design of BP Speaker Recognition System Based on KPCA-MFCC Parameter Optimization[C]// 2018 International Conference on Mechanical, Electrical, Electronic Engineering & Science (MEEES 2018). 0.

[7] Wang H, Hu W. Optimization of Pathological Voice Feature Based on KPCA and SVM[C]// Chinese Conference on Biometric Recognition. Springer International Publishing, 2014.

[8] 2021.Chunlei Yu, Mengyue Li , Weishi Yin . Journal of Changchun University of Science and Technology: Natural Science Edition, 2021.

[9] Xiaoling Chen. Stock price prediction based on ARIMA model and neural network model [J]. Economic Mathematics, 2017, 34(4):5.

[10] Peter G, Zhang. Time series forecasting using a hybrid ARIMA and neural network model [J]. Neurocomputing, 2003.

[11] Ümit Çavuş Büyükşahin, Şeyda Ertekin. Improving forecasting accuracy of time series data using a new ARIMA-ANN hybrid method and empirical mode decomposition [J]. Neurocomputing, 2019, 361:151-163.

[12] Xinyu Zhang, Guohua Zou. Model average method and its application in forecasting, Statistical Research, 2011, 28(6):6.