본문

서브메뉴

Topics on Machine Learning in Finance
Topics on Machine Learning in Finance
Topics on Machine Learning in Finance

상세정보

자료유형  
 학위논문 서양
최종처리일시  
20260202103506
ISBN  
9798280748101
DDC  
310
저자명  
Zheng, Yuheng.
서명/저자  
Topics on Machine Learning in Finance
발행사항  
[Sl] : Princeton University, 2025
발행사항  
Ann Arbor : ProQuest Dissertations & Theses, 2025
형태사항  
386 p
주기사항  
Source: Dissertations Abstracts International, Volume: 86-12, Section: B.
주기사항  
Advisor: Ait-Sahalia, Yacine;Fan, Jianqing.
학위논문주기  
Thesis (Ph.D.)--Princeton University, 2025.
초록/해제  
요약In recent years, artificial intelligence (AI) and machine learning (ML) have become essential tools in financial markets, with applications spanning from high-frequency trading to risk management. This thesis explores the theoretical foundations and practical implications of ML-driven methods in financial decision-making, focusing on reinforcement learning, statistical factor models, and diffusion models with latent factors.In Chapter 2, we establish a rigorous theoretical framework for reinforcement learning (RL) in high-frequency market making, bridging modern RL theory with continuous-time financial models. We analyze the effect of sampling frequency and uncover a tradeoff between learning error and sample complexity. Extending to a multi-agent setting, we study a two-player general-sum game and show that Nash equilibria of the discrete-time RL framework converge to their continuous-time counterparts as ∆ → 0. Our findings provide guidance for practitioners selecting sampling frequencies in high-frequency financial decision-making. In Chapter 3, we develop an estimation and inference theory for principal component analysis (PCA) under a weak factor model with cross-sectionally dependent idiosyncratic components, with the nearly minimal factor strength relative to the noise level or signal-to-noise ratio. Our theoretical innovations enable novel statistical tests for testing latent factor space, detecting structural breaks, and estimating systematic risks, with empirical validation revealing strong correlations between our test results and macroeconomic cycles.In Chapter 4, we establish the asymptotic behavior of marginal-information and full-information maximum likelihood estimators for discretely-sampled continuous-time diffusion models with latent factors. Our results demonstrate that the cost of latency, which we define as the amount of information lost when a variable is latent instead of observed, depends on whether the latent factors affect the drift or diffusion terms. To obtain these results, we employ filtering techniques and develop a new theory of nonlinear continuous-time pseudo-filters based on stochastic partial differential equations (SPDE). Monte Carlo simulations provide empirical support for our theoretical findings on the cost of latency.
일반주제명  
Statistics
일반주제명  
Finance
일반주제명  
Mathematics
키워드  
Diffusion models
키워드  
Factor model
키워드  
Machine learning
키워드  
Market making
키워드  
Nonlinear filtering
키워드  
Reinforcement learning
기타저자  
Princeton University Operations Research and Financial Engineering
기본자료저록  
Dissertations Abstracts International. 86-12B.
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008260126s2025        us                              c    eng  d
■001000017357398
■00520260202103506
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798280748101
■035    ▼a(MiAaPQ)AAI32002703
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a310
■1001  ▼aZheng,  Yuheng.▼0(orcid)0009-0006-8572-7365
■24510▼aTopics  on  Machine  Learning  in  Finance
■260    ▼a[Sl]▼bPrinceton  University▼c2025
■260  1▼aAnn  Arbor▼bProQuest  Dissertations  &  Theses▼c2025
■300    ▼a386  p
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  86-12,  Section:  B.
■500    ▼aAdvisor:  Ait-Sahalia,  Yacine;Fan,  Jianqing.
■5021  ▼aThesis  (Ph.D.)--Princeton  University,  2025.
■520    ▼aIn  recent  years,  artificial  intelligence  (AI)  and  machine  learning  (ML)  have  become  essential  tools  in  financial  markets,  with  applications  spanning  from  high-frequency  trading  to  risk  management.  This  thesis  explores  the  theoretical  foundations  and  practical  implications  of  ML-driven  methods  in  financial  decision-making,  focusing  on  reinforcement  learning,  statistical  factor  models,  and  diffusion  models  with  latent  factors.In  Chapter  2,  we  establish  a  rigorous  theoretical  framework  for  reinforcement  learning  (RL)  in  high-frequency  market  making,  bridging  modern  RL  theory  with  continuous-time  financial  models.  We  analyze  the  effect  of  sampling  frequency  and  uncover  a  tradeoff  between  learning  error  and  sample  complexity.  Extending  to  a  multi-agent  setting,  we  study  a  two-player  general-sum  game  and  show  that  Nash  equilibria  of  the  discrete-time  RL  framework  converge  to  their  continuous-time  counterparts  as  ∆  →  0.  Our  findings  provide  guidance  for  practitioners  selecting  sampling  frequencies  in  high-frequency  financial  decision-making. In  Chapter  3,  we  develop  an  estimation  and  inference  theory  for  principal  component  analysis  (PCA)  under  a  weak  factor  model  with  cross-sectionally  dependent  idiosyncratic  components,  with  the  nearly  minimal  factor  strength  relative  to  the  noise  level  or  signal-to-noise  ratio.  Our  theoretical  innovations  enable  novel  statistical  tests  for  testing  latent  factor  space,  detecting  structural  breaks,  and  estimating  systematic  risks,  with  empirical  validation  revealing  strong  correlations  between  our  test  results  and  macroeconomic  cycles.In  Chapter  4,  we  establish  the  asymptotic  behavior  of  marginal-information  and  full-information  maximum  likelihood  estimators  for  discretely-sampled  continuous-time  diffusion  models  with  latent  factors.  Our  results  demonstrate  that  the  cost  of  latency,  which  we  define  as  the  amount  of  information  lost  when  a  variable  is  latent  instead  of  observed,  depends  on  whether  the  latent  factors  affect  the  drift  or  diffusion  terms.  To  obtain  these  results,  we  employ  filtering  techniques  and  develop  a  new  theory  of  nonlinear  continuous-time  pseudo-filters  based  on  stochastic  partial  differential  equations  (SPDE).  Monte  Carlo  simulations  provide  empirical  support  for  our  theoretical  findings  on  the  cost  of  latency.
■590    ▼aSchool  code:  0181.
■650  4▼aStatistics
■650  4▼aFinance
■650  4▼aMathematics
■653    ▼aDiffusion  models
■653    ▼aFactor  model
■653    ▼aMachine  learning
■653    ▼aMarket  making
■653    ▼aNonlinear  filtering
■653    ▼aReinforcement  learning
■690    ▼a0463
■690    ▼a0508
■690    ▼a0405
■690    ▼a0800
■71020▼aPrinceton  University▼bOperations  Research  and  Financial  Engineering.
■7730  ▼tDissertations  Abstracts  International▼g86-12B.
■790    ▼a0181
■791    ▼aPh.D.
■792    ▼a2025
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17357398▼nKERIS▼z이  자료의  원문은  한국교육학술정보원에서  제공합니다.

미리보기

내보내기

chatGPT토론

Ai 추천 관련 도서


    신착도서 더보기
    최근 3년간 통계입니다.

    소장정보

    • 예약
    • 소재불명신고
    • 나의폴더
    • 우선정리요청
    • 비도서대출신청
    • 야간 도서대출신청
    소장자료
    등록번호 청구기호 소장처 대출가능여부 대출정보
    TF17401 전자도서 대출가능 마이폴더 부재도서신고 비도서대출신청 야간 도서대출신청

    * 대출중인 자료에 한하여 예약이 가능합니다. 예약을 원하시면 예약버튼을 클릭하십시오.

    해당 도서를 다른 이용자가 함께 대출한 도서

    관련 인기도서

    로그인 후 이용 가능합니다.