본문 바로가기
KERD
논문KCI 등재

장기 변동성을 반영하기 위한 확산 모델을 활용한 주가 지수 확률 분포 예측

Stock Index Probabilistic Distribution Forecasting Using a Diffusion Model to Reflect the Volatility of Long-Term Sequences

김영진|김하영
한국지식정보기술학회 (Korea Knowledge Information Technology Society)|2023.02|한국지식정보기술학회 논문지 (Journal of Knowledge Information Technology and Systems)|제18권 제1호|pp. 141-155|DOI 10.34163/jkits.2023.18.1.015

0

조회

0

다운로드

3

인용

0

참고문헌

키워드

Deep learningTime series forecastingStock market indexStochastic differential equationsFinancial time seriesProbabilistic forecasting
주제분류학제간연구

초록

기존 주식 시장의 시계열 예측 연구의 상당수는 점 예측을 기반으로 한 통계적 또는 기계 학습 방법론이다. 그러나 일반적인 점 예측 방법은 데이터의 분포를 고려하지 못해 극단적인 시장 상황을 무시할 수 있다. 이는 주식 시장의 높은 변동성을 감안할 때, 장기 예측에 취약할 수 있다는 한계점을 갖는다. 이 문제를 해결하기 위해, 우리는 디노이징 확산 모델을 기반으로 한 주가 지수 확률 분포 예측 프레임워크인 StockGrad를 제안한다. 구체적으로 StockGrad는 예측 기간과 관련성이 낮은 기간에 대한 확률을 0으로 산출하기 위해 희소 트랜스포머 인코더를 특징 추출기로 채택한다. 확률 분포 예측을 위해, 본 프레임워크는 TimeGrad 시계열 생성 방법론을 사용하여 마르코프 체인을 통해 백색 잡음을 관심 분포로 변환한다. 우리가 제안한 방법론은 투자자들이 투자 또는 위험 관리에 대한 결정을 내리는 데 도움이 될 수 있다. 실험에 따르면 StockGrad는 S&P500, NIKKEI225, KOSPI200 주가 지수 데이터 세트에서 기존 딥러닝 시계열 확률 모델보다 각각 약 3.71%, 1.53%, 1.71% 성능이 좋게 나옴을 확인할 수 있다. 또한, 희소 트랜스포머 인코더를 특징 추출기로 사용하면 예측 구간과 관련된 과거 시점을 잘 포착하게 되어 장기 예측 성능이 향상된다는 것이 실험적으로 입증되었다.

Abstract

Many of the existing time series forecasting studies in stock market were statistical or machine learning methods based on point prediction. However, typical point forecasting methods fail to consider the distribution of data, which may ignore extreme market situations. This can be vulnerable especially for the long-term forecasting, given the high volatility in the stock market. To resolve the issue, we propose a stock index probabilistic distribution prediction framework based on the Denoising Diffusion Model, called StockGrad. Specifically, StockGrad adopt a sparse transformer encoder as feature extractor to yield zero probability for time steps that are less relevant to the prediction time step. For probability distribution prediction, our model transforms the white noise into the distribution of interest through the Markov chain using the TimeGrad time series generation methodology. Our proposed methodology can help investors make decisions on investment or risk management. Our experiments show that proposed StockGrad framework outperforms the existing deep learning probability models by about 3.71%, 1.53% and 1.71% on S&P500, NIKKEI225, and KOSPI200 stock index data sets, respectively. Also, it is experimentally demonstrated that using the sparse transformer encoder as a feature extractor captures historical time points related to predictive bulbs well, improve the performance of the long-term prediction.

참고문헌 (39)

  1. K. Miao, "Stock price forecast based on bacterial colony RBF neural network", Journal of Qingdao University(Natural Science Edition), 2(11) (2007)
  2. T. Muhammad, "Transformer-Based deep learning model for stock price prediction : A case study on bangladesh stock market", arXiv preprint arXiv : 2208. 08300, ~ (2022)
  3. T. Bollerslev, "Stock return predictability and variance risk premia : statistical inference and international evidence", Journal of Financial and Quantitative Analysis, 49(3), pp.633~661 (2014)
  4. D. H. B. Phan, "Stock return forecasting : Some new evidence", International Review of Financial Analysis, 40, pp.38~51 (2015)
  5. D. Önkal, "Evaluating probabilistic forecasts of stock prices in a developing stock market", European Journal of Operational Research, 74(2), pp.350~358
  6. S. Wu, "Adversarial sparse transformer for time series forecasting", Advances in neural information processing systems, 33, pp.17105~17115 (2020)
  7. D. Salinas, "DeepAR: Probabilistic forecasting with autoregressive recurrent networks", International Journal of Forecasting, 36(3), pp.1181~1191 (2020)
  8. C. Winkler, "Learning likelihoods with conditional normalizing flows", arXiv preprint arXiv : 1912. 00042, ~ (2019)
  9. D. E. Rumelhart, "Learning representations by back-propagating errors", nature, 323(6088), pp.533~536 (1986)
  10. A. Vaswani, "Attention is all you need", Advances in neural information processing systems, 30 (2017)
  11. S. Hochreiter, "Long short-term memory", Neural computation, 9(8), pp.1735~1780 (1997)
  12. J. Chung, "Empirical evaluation of gated recurrent neural networks on sequence modeling", arXiv preprint arXiv : 1412. 3555, ~ (2014)
  13. R. C. Staudemeyer, "Understanding LSTM--a tutorial into long short-term memory recurrent neural networks", arXiv preprint arXiv : 1909. 09586, ~ (2019)
  14. R. Child, "Generating long sequences with sparse transformers", arXiv preprint arXiv : 1904. 10509, ~ (2019)
  15. J. Ho, "Denoising diffusion probabilistic models", Advances in Neural Information Processing Systems, 33, pp.6840~6851 (2020)
  16. K. Rasul, "Autoregressive denoising diffusion models for multivariate probabilistic time series forecasting", International Conference on Machine Learning(ICML) (2021), PMLR
  17. G. E. P. Box, "Distribution of residual autocorrelations in autoregressive-integrated moving average time series models", Journal of the American statistical Association, 65(332), pp.1509~1526 (1970)
  18. T. Bollerslev, "Generalized autoregressive conditional heteroskedasticity", Journal of econometrics, 31(3), pp.307~327 (1986)
  19. F. M. Bianchi, "An overview and comparative analysis of recurrent neural networks for short term load forecasting", arXiv preprint arXiv : 1705. 04378, ~ (2017)
  20. M. Längkvist, "A review of unsupervised feature learning and deep learning for time-series modeling", Pattern Recognition Letters, 42, pp.11~24 (2014)
  21. K. Rasul, "Multivariate probabilistic time series forecasting via conditioned normalizing flows", arXiv preprint arXiv : 2002. 06103, ~ (2020)
  22. A. Hossain, "Comparison of the finite mixture of ARMA-GARCH, back propagation neural networks and support-vector machines in forecasting financial returns", Journal of Applied Statistics, 38(3), pp.533~551 (2011)
  23. W. Wang, "Stock indices analysis based on ARMA-GARCH model", ieee international conference on industrial engineering and engineering management. IEEE (2009), IEEE
  24. 유희경, "일반화 자기회귀 조건부 이분산 모형들의 베이지안 군집분석에 관한 연구: 주식자료에 응용", 한국지식정보기술학회 논문지, 7(6), pp.245~254 (2012), 한국지식정보기술학회
  25. L. I. Zhi-Peng, "An improved adaptive exponential smoothing model for short-term travel time forecasting of urban arterial street", Acta automatica sinica, 34(11), pp.1404~1409 (2008)
  26. R. Hecht-Nielsen, "Theory of the backpropagation neural network, Neural networks for perception", pp.65~93 (1992), Academic Press
  27. J. J. Wang, "Stock index forecasting based on a hybrid model", Omega, 40(6), pp.758~766 (2012)
  28. J. H. Friedman, "Multivariate adaptive regression splines", The annals of statistics, 19(1), pp.1~67 (1991)
  29. A. J. Smola, "A tutorial on support vector regression", Statistics and computing, 14(3), pp.199~222 (2004)
  30. L. J. Kao, "A hybrid approach by integrating wavelet-based feature extraction with MARS and SVR for stock index forecasting", Decision Support Systems, 54(3), pp.1228~1244 (2013)
  31. J. Wang, "Forecasting stock market indexes using principle component analysis and stochastic time effective neural networks", Neurocomputing, 156, pp.68~78 (2015)
  32. 송한진, "AI의 LSTM기법을 이용한 금융시계열 데이터 변동성 예측방법 연구", 한국지식정보기술학회 논문지, 14(6), pp.665~673 (2019), 한국지식정보기술학회[DOI]
  33. Z. Chengzaho, "Analysis of Asia Pacific stock markets with a novel multiscale model", Physica A, 534(120939) (2019)
  34. H. Niu, "A hybrid stock price index forecasting model based on variational mode decomposition and LSTM network", Applied Intelligence, 50(12), pp.4296~4309 (2020)
  35. Y. J. Baek, "ModAugNet : A new forecasting framework for stock market index value with an overfitting prevention LSTM module and a prediction LSTM module", Expert Systems with Applications, 113, pp.457~480 (2018)
  37. J. E. Matheson, "Scoring rules for continuous probability distributions", Management science, 22(10), pp.1087~1096 (1976)
  38. A. Jordan, "Evaluating probabilistic forecasts with scoringRules", arXiv preprint arXiv : 1709. 04743, ~ (2017)
  39. D. Salinas, "High-dimensional multivariate forecasting with low-rank gaussian copula processes", Advances in neural information processing systems, 32 (2019)

관련 논문AI추천

KOSPI200 Index Prediction Using Sequence-to-Sequence based on Denoising Filter and Attention Mechanism오경주, 이상규, 이희수자연과학연구소 · Quantitative Bio-Science2020 · 제39권 제2호해외지수와 투자자별 매매 동향에 따른 딥러닝 기반 주가 등락 예측김태승, 이수원한국정보처리학회 · 정보처리학회 논문지2021 · 제10권 제9호Forecasting the KOSPI 200 Stock Index Based on LSTM Autoencoder김영민, 양준혁, 오경주자연과학연구소 · Quantitative Bio-Science2020 · 제39권 제2호Prediction of KOSPI 200 Index Movement Using Generative Adversarial Network Based on LSTM오경주, 주성준자연과학연구소 · Quantitative Bio-Science2023 · 제42권 제2호TimeGAN을 활용한 트레이딩 알고리즘 선택송재원, 이재윤, 이주홍, 최범기(사)한국스마트미디어학회 · 스마트미디어저널2022 · 제11권 제1호KOSPI200 Prediction through Low-Pass Filtered Long Short-Term Memory Algorithm오경주, 이동원, 이희수자연과학연구소 · Quantitative Bio-Science2020 · 제39권 제1호

이 간행물의 다른 논문 (22건)

딥러닝 알고리즘 기반의 자동차 현가장치 부품 수요 예측 모형에 대한 연구이상길, 김준우pp. 1-13Real-Time Processing System of E-Commerce User Data Based on Spark Streaming장도, 가문초, 김은성, 정회경pp. 15-24다양한 탐지 알고리즘을 이용한 탐지 데이터셋에 관한 연구이세열pp. 25-30ChatGPT를 이용한 문헌 작성 설계 및 이슈박정훈, 정한민pp. 31-40한국 리셀 플랫폼 활성화 방안김광현, 김원빈pp. 41-48
원문 보기