Recurrent neural network-adapted nonlinear ARMA-GARCH model with application to S&P 500 index data †
Yongjin Jeong1 · Sangyeol Lee2
12 Department of Statistics, Seoul National University
Correspondence to: Professor, Department of Statistics, Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 08826, South Korea. E-mail: sylee@stats.snu.ac.kr
† This work was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (No. 2018R1A2A2A05019433).
† This work was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (No. 2018R1A2A2A05019433).
Received August 13, 2019; Revised September 2, 2019; Accepted September 11, 2019.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
- Abstract
- This study designs a nonlinear ARMA-GARCH model well adapted to a recurrent neural network (RNN) scheme. The classical ARMA-GARCH model has been a func- tional device to model financial time series with high volatilities. However, its linear structure somewhat restricts its usage in practice because of nonlinear and nonstation- ary features of time series. Considering this, we incorporate hyperbolic tangent func- tions into the ARMA component for improving the adaptability to RNN, and suggest an RNN-adapted nonlinear ARMA-GARCH model. This model is evaluated through a comparison study with the linear ARMA-GARCH model by applying algorithmic trad- ing strategies and forecasts to the S&P500 daily closing index. Our findings confirm the validity of the proposed method.
- Keywords : ARMA-GARCH model, recurrent neural network, S&P500 index, time series forecasting.
Full Text(PDF) Free
