View : 651 Download: 0

Full metadata record

DC Field Value Language
dc.contributor.advisor신동완-
dc.contributor.author서영은-
dc.creator서영은-
dc.date.accessioned2021-01-25T16:30:23Z-
dc.date.available2021-01-25T16:30:23Z-
dc.date.issued2021-
dc.identifier.otherOAK-000000172908-
dc.identifier.urihttp://dcollection.ewha.ac.kr/common/orgView/000000172908en_US
dc.identifier.urihttps://dspace.ewha.ac.kr/handle/2015.oak/256118-
dc.description.abstractForecasting stock volatility is made using autoregressive (AR) model, vector autoregressive (VAR) model and deep learning methods of the recurrent neural network (RNN): simple recurrent neural network (simple RNN), gated recurrent unit (GRU), long short term memory (LSTM). We consider two US and one Korea stock indices: DJIA, SPX, KS11. And we forecast realized volatilities of 5-minute log returns using online search engine data from Google and Naver. An out-of-sample forecast comparison shows that RNN model performs better than the other time series models. Also, we investigate the influence of using search volume data on each stock volatility forecasts. The model using search volume improves forecasts on DJI and KS11.;본 연구는 자기회귀(Autoregressive; AR) 모델, 벡터자기회귀 (Vector Autoregressive; VAR) 모델 및 순환 신경망(Recurrent Neural Network; RNN) 방법을 사용하여 주식 변동성을 예측한다. RNN은 단순 순환 신경망(Simple RNN), 게이트 순환 유닛(Gated Recurrent Unit; GRU), 장기 기억장치(Long-Short Term Memory; LSTM) 셀(Cell)을 사용했다. 미국과 한국의 3가지 주식 지수 DJIA, SPX, KS11가 고려되었고, 구글, 네이버의 온라인 검색엔진 데이터를 사용했다. 표본 외 예측 비교를 통해 RNN 모형은 다른 시계열 모델보다 성능이 우수하다는 것을 알 수 있다. 또한 주가 변동성 예측에 대해 검색량 데이터의 영향을 살펴봤을 때, 검색 볼륨을 사용한 모델은 DJI, KS11에서 예측력이 개선된다.-
dc.description.tableofcontentsI. Introduction 1 II. Data 3 A. Realized volatility data 3 B. Search volume data 3 III. Model Description 8 A. Autoregressive (AR) model 8 B. Vector Autoregressive (VAR) model 8 C. Recurrent Neural Networks (RNNs) 9 IV. Experiment and Results 13 A. Data separation and normalization 13 B. Hyperparameter determination 14 C. Model Evaluation 15 D. Results. 16 V. Conclusion 22 Bibliography 23 Abstract(in Korean) 25-
dc.formatapplication/pdf-
dc.format.extent1341384 bytes-
dc.languageeng-
dc.publisher이화여자대학교 대학원-
dc.subject.ddc500-
dc.titleForecasting Stock Volatility Using Deep Learning Methods with Search Queries-
dc.typeMaster's Thesis-
dc.creator.othernameSeo, Yeong eun-
dc.format.pageiv, 25 p.-
dc.identifier.thesisdegreeMaster-
dc.identifier.major대학원 통계학과-
dc.date.awarded2021. 2-
Appears in Collections:
일반대학원 > 통계학과 > Theses_Master
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

BROWSE