Search

Stock Index Forecasting of CSI 300 Based on a Hybrid ARIMA-LSTM Model

Author(s)

Yunhao Lin

Affiliation(s)

Department of Applied Mathematics with Economics, Jinan University, Guangzhou, Guangdong, China *Corresponding Author

Abstract

As a key barometer of China’s capital markets, the CSI 300 Index holds significant value for investment decision-making and risk management. In addressing the limitations of conventional forecasting models in processing nonlinear and non-stationary financial time series data, this study proposes a hybrid prediction framework that integrates Autoregressive Integrated Moving Average (ARIMA), Variational Mode Decomposition (VMD), and Long Short-Term Memory (LSTM) networks. The proposed model employs a multi-stage strategy involving signal decomposition, feature modelling, and result integration to comprehensively capture both the linear trends and nonlinear fluctuations in time series data, thereby enhancing forecasting accuracy and generalization capability. Based on historical CSI 300 data, the study compares forecasting performance across different time horizons. Results indicate that, compared with standalone ARIMA and LSTM models, the hybrid model demonstrates superior stability and adaptability in short-, medium-, and long-term forecasting. Notably, in long-term scenarios, the integration of multi-scale features effectively mitigates error accumulation, confirming the robustness of the hybrid approach under complex market conditions.

Keywords

CSI 300 Index; Time Series Forecasting; ARIMA Model; Variational Mode Decomposition (VMD); Long Short-Term Memory (LSTM); Hybrid Prediction Model

References

[1] Ci, B., Zhang, P. (2022) Financial time series prediction based on ARIMA-LSTM model. Stat. Decis., 38(11): 145–149. [2] Guo, J. (2020) Research on CSI 300 index prediction based on VMD-EEMD-LSTM model. Mod. Finance (Tianjin Univ. Finance Econ. J.), 40(08): 31–44. [3] Lei, K., Chen, Y. (2007) Prediction of China's inbound tourist volume based on BP neural network and ARIMA hybrid model. J. Tourism, (04): 20–25. [4] Wu, Y., Wen, X. (2016) Short-term stock price prediction based on ARIMA model. Stat. Decis., (23): 83–86. [5] Zhang, L., Sun, S., Wang, Y. (2021) Exchange rate prediction based on deep learning LSTM model. Stat. Decis., 37(13): 158–162. [6] Zhou, Z., He, X. (2023) Stock price prediction method based on optimized LSTM model. Stat. Decis., 39(06): 143–148. [7] Ariyo, A.A., Adewumi, A.O., Ayo, C.K. (2014) Stock price prediction using the ARIMA model. Procedia Comput. Sci. [8] Dave, E., Leonardo, A., Jeanice, M., et al. (2021) Forecasting Indonesia exports using a hybrid model ARIMA-LSTM. Procedia Comput. Sci., 179: 480–487. [9] Pierre, A.A., Akim, S.A., Semenyo, A.K., et al. (2023) Peak electrical energy consumption prediction by ARIMA, LSTM, GRU, ARIMA-LSTM and ARIMA-GRU approaches. [10] Wen, T., Liu, Y., Bai, Y.H., et al. (2023) Modeling and forecasting CO₂ emissions in China and its regions using a novel ARIMA-LSTM model. Heliyon, 9(11): e2023. [11] Wu, J., Xu, K., Chen, X., et al. (2021) Price graphs: Utilizing the structural information of financial time series for stock prediction. Papers.