DOI QR코드

DOI QR Code

Forecasting volatility index by temporal convolutional neural network

Causal temporal convolutional neural network를 이용한 변동성 지수 예측

  • Ji Won Shin (Institute of Mathematical Sciences, Ewha Womans University) ;
  • Dong Wan Shin (Department of Statistics, Ewha Womans University)
  • 신지원 (이화여자대학교 수리과학연구소) ;
  • 신동완 (이화여자대학교 통계학과)
  • Received : 2022.12.14
  • Accepted : 2022.12.31
  • Published : 2023.04.30

Abstract

Forecasting volatility is essential to avoiding the risk caused by the uncertainties of an financial asset. Complicated financial volatility features such as ambiguity between non-stationarity and stationarity, asymmetry, long-memory, sudden fairly large values like outliers bring great challenges to volatility forecasts. In order to address such complicated features implicity, we consider machine leaning models such as LSTM (1997) and GRU (2014), which are known to be suitable for existing time series forecasting. However, there are the problems of vanishing gradients, of enormous amount of computation, and of a huge memory. To solve these problems, a causal temporal convolutional network (TCN) model, an advanced form of 1D CNN, is also applied. It is confirmed that the overall forecasting power of TCN model is higher than that of the RNN models in forecasting VIX, VXD, and VXN, the daily volatility indices of S&P 500, DJIA, Nasdaq, respectively.

변동성의 예측은 자산의 리스크에 대비하는 데에 중요한 역할을 하기때문에 필수적이다. 인공지능을 통하여 이러한 복잡한 특성을 지닌 변동성 예측을 시도하였는데 기존 시계열 예측에 적합하다 알려진 LSTM (1997)과 GRU (2014)은 기울기 소실로 인한 문제, 방대한 연산량의 문제, 그로 인한 메모리양의 문제 등이 존재하였다. 변동성 데이터는 비정상성(non-stationarity)과 정상성(stationarity)을 모두 가지고 있는 특성이 있으며, 자산 가격 하방 쇼크에 더 큰 폭으로 상승하는 비대칭성과 상당한 장기 기억성, 시장에 큰 사건이 발생할 때 기존의 값들에 비해 이상치라 할 수 있을 정도의 예측할 수 없는 큰 값이 발생하는 특성들이 존재한다. 이렇게 여러 가지 복잡한 특성들은 하나의 모형으로 구조화되기 어려워서 전통적인 방식의 모형으로는 변동성에 대한 예측력을 높이기 어려운 면이 있다. 이러한 문제를 해결하기 위해 1D CNN의 발전된 형태인 causal TCN (causal temporal convolutional network) 모형을 변동성 예측에 적용하고, 예측력을 최대화 할 수 있는 TCN 구조를 설계하고자 하였다. S&P 500, DJIA, Nasdaq 지수에 해당하는 변동성 지수 VIX, VXD, and VXN, 에 대하여 예측력 비교를 하였으며, TCN 모형이 RNN 계열의 모형보다도 전반적으로 예측력이 높음을 확인하였다.

Keywords

Acknowledgement

이 성과는 한국 연구재단의 지원을 받아 수행된 연구임 (No. 2019R1A6A1A11051177).

References

  1. Bai S, Kolter JZ, and Koltun V (2018). An empirical evaluation of generic convolutional and recurrent networks for sequence modeling, Available from: arXiv preprint arXiv: 1803.01271
  2. Bi J, Zhang X, Yuan H, Zhang J, and Zhou M (2021). A hybrid prediction method for realistic network traffic with temporal convolutional network and lstm, IEEE Transactions on Automation Science and Engineering, 19, 1869-1879. https://doi.org/10.1109/TASE.2021.3077537
  3. Bucci A (2017). Forecasting realized volatility: A review, Journal of Advanced Studies in Finance, 8, 94-138.
  4. Chou J and Ngom N (2016). Time series analytics using sliding window metaheuristic optimization-based machine learning system for identifying building energy consumption patterns, Applied Energy, 177, 751-770. https://doi.org/10.1016/j.apenergy.2016.05.074
  5. Chung J, Gulcehre C, Cho KH, and Bengio Y (2014). Empirical Evaluation of Gated Recurrent Neural Network-son Sequence Modeling, Available from: arXiv: :1412.3555
  6. Hansen PR, Lunde A, and Nason JM (2011). The model confidence set, Econometrica, 79, 453-497. https://doi.org/10.3982/ECTA5771
  7. Hewage P, Behera A, Trovati M, Pereira E, Ghahremani M, Palmieri F, and Liu Y (2020). Temporal convolutional neural (TCN) network for an effective weather forecasting using time-series data from the local weather station, Soft Computing, 24, 16453-16482. https://doi.org/10.1007/s00500-020-04954-0
  8. Hochreiter S and Schmidhuber J (1997). Long short-term memory, Neural Computation, 9, 1735-1780. https://doi.org/10.1162/neco.1997.9.8.1735
  9. Lara-Benitez P, Carranza-Garcia M, Luna-Romera JM, and Riquelme JC (2020). Temporal convolutional networks applied to energy-related time series forecasting, Applied Sciences, 10, 2322.
  10. Lee D, Lee S, Han Y, and Lee K (2017). Ensemble of convolutional neural networks for weakly-supervised soundevent detection using multiple scale input, Detection and Classification of Acoustic Scenes and Events, 2017, 14-18.
  11. Li W, Wei Y, An D, Jiao Y, and Wei Q (2022). LSTM-TCN: Dissolved oxygen prediction in aquaculture, based on combined model of long short-term memory network and temporal convolutional network, Environmental Science and Pollution Research, 29, 39545-39556. https://doi.org/10.1007/s11356-022-18914-8
  12. Liu Q, Che X, and Bie M (2019). R-STAN: Residual spatial-temporal attention network for action recognition, IEEE Access, 7, 82246-82255. https://doi.org/10.1109/ACCESS.2019.2923651
  13. McAleer M and Medeiros MC (2008). A multiple regime smooth transition heterogeneous autoregressive modelfor long memory and asymmetries, Journal of Econometrics, 147, 104-119. https://doi.org/10.1016/j.jeconom.2008.09.032
  14. Poon SH and Granger CWJ (2003). Forecasting volatility in financial markets: A review, Journal of EconomicLiterature, 41, 478-539. https://doi.org/10.1257/jel.41.2.478
  15. Shin JW and Shin DW (2019). Vector error correction heterogeneous autoregressive forecast model of realized volatility and implied volatility, Communications in Statistics-Simulation and Computation, 48, 1503-1515. https://doi.org/10.1080/03610918.2017.1414250
  16. Shin JW and Shin DW (2022). Deep learning forecasting for financial realized volatilitieswith aid of implied volatilities and internet search volumes, The Korean Journal of Applied Statistics, 35, 93-104.  https://doi.org/10.5351/KJAS.2022.35.1.093