서브메뉴
검색
Trustworthy and Robust Early Sepsis Prediction for Intensive Care Unit Patients Using Reinforcement Learning and Conformal Prediction
Trustworthy and Robust Early Sepsis Prediction for Intensive Care Unit Patients Using Reinforcement Learning and Conformal Prediction
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105526
- ISBN
- 9798263341015
- DDC
- 616.028
- 저자명
- Zhou, Anni.
- 서명/저자
- Trustworthy and Robust Early Sepsis Prediction for Intensive Care Unit Patients Using Reinforcement Learning and Conformal Prediction
- 발행사항
- [Sl] : Georgia Institute of Technology, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 141 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-05, Section: A.
- 주기사항
- Advisor: Kamaleswaran, Rishikesan;Beyah, Raheem A.
- 학위논문주기
- Thesis (Ph.D.)--Georgia Institute of Technology, 2025.
- 초록/해제
- 요약This thesis presents a comprehensive series of studies aimed at improving early sepsis prediction for ICU patients using advanced machine learning frameworks and algorithms. In chapter 1, we introduce the background of early sepsis prediction problems, challenges of existing machine learning approaches applied in this scenario, and the motivations of utilizing Multi-armed Bandit (MAB) and Conformal Prediction in our studies.The first study introduced OnAI-Comp, a novel multi-armed bandit-based framework where a group of AI experts competes, and the best-performing expert is selected for each patient. This approach enables adaptive decision-making and predicts the sepsis onset time by leveraging the diversity of expert models and incorporating patient-specific contexts.The second study developed Sepsyn-OLCP, a reinforcement learning algorithm integrated with conformal prediction to ensure the reliability of its predictions. By combining the robustness of gap-based bandits with the statistical guarantees of conformal prediction, this method delivers predictions that are not only accurate but also trustworthy, addressing a critical need in high-stakes healthcare applications.The third study advanced the state of the art by proposing NeuroSep-CP-LCB, a cutting-edge algorithm that fuses neural network-based contextual bandits with conformal prediction methodologies. This approach employs neural networks to approximate the reward function with high fidelity, allowing for more nuanced and data-driven decisions. The integration of conformal prediction provides calibrated prediction intervals, ensuring the trustworthiness of the results. By incorporating principles of contextual bandits, the algorithm efficiently balances exploration and exploitation, ensuring optimal decision-making under uncertainty.These studies contribute to the field of predictive modeling for sepsis, offering frameworks and algorithms that prioritize both predictive accuracy and reliability, crucial for critical care environments.
- 일반주제명
- Intensive care
- 일반주제명
- Distance learning
- 일반주제명
- Mortality
- 일반주제명
- Decision making
- 일반주제명
- Neural networks
- 일반주제명
- Educational technology
- 기본자료저록
- Dissertations Abstracts International. 87-05A.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017360438
■00520260202105526
■006m o d
■007cr#unu||||||||
■020 ▼a9798263341015
■035 ▼a(MiAaPQ)AAI32309764
■035 ▼a(MiAaPQ)GeorgiaTech77776
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a616.028
■1001 ▼aZhou, Anni.
■24510▼aTrustworthy and Robust Early Sepsis Prediction for Intensive Care Unit Patients Using Reinforcement Learning and Conformal Prediction
■260 ▼a[Sl]▼bGeorgia Institute of Technology▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a141 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-05, Section: A.
■500 ▼aAdvisor: Kamaleswaran, Rishikesan;Beyah, Raheem A.
■5021 ▼aThesis (Ph.D.)--Georgia Institute of Technology, 2025.
■520 ▼aThis thesis presents a comprehensive series of studies aimed at improving early sepsis prediction for ICU patients using advanced machine learning frameworks and algorithms. In chapter 1, we introduce the background of early sepsis prediction problems, challenges of existing machine learning approaches applied in this scenario, and the motivations of utilizing Multi-armed Bandit (MAB) and Conformal Prediction in our studies.The first study introduced OnAI-Comp, a novel multi-armed bandit-based framework where a group of AI experts competes, and the best-performing expert is selected for each patient. This approach enables adaptive decision-making and predicts the sepsis onset time by leveraging the diversity of expert models and incorporating patient-specific contexts.The second study developed Sepsyn-OLCP, a reinforcement learning algorithm integrated with conformal prediction to ensure the reliability of its predictions. By combining the robustness of gap-based bandits with the statistical guarantees of conformal prediction, this method delivers predictions that are not only accurate but also trustworthy, addressing a critical need in high-stakes healthcare applications.The third study advanced the state of the art by proposing NeuroSep-CP-LCB, a cutting-edge algorithm that fuses neural network-based contextual bandits with conformal prediction methodologies. This approach employs neural networks to approximate the reward function with high fidelity, allowing for more nuanced and data-driven decisions. The integration of conformal prediction provides calibrated prediction intervals, ensuring the trustworthiness of the results. By incorporating principles of contextual bandits, the algorithm efficiently balances exploration and exploitation, ensuring optimal decision-making under uncertainty.These studies contribute to the field of predictive modeling for sepsis, offering frameworks and algorithms that prioritize both predictive accuracy and reliability, crucial for critical care environments.
■590 ▼aSchool code: 0078.
■650 4▼aIntensive care
■650 4▼aDistance learning
■650 4▼aMortality
■650 4▼aDecision making
■650 4▼aNeural networks
■650 4▼aEducational technology
■690 ▼a0800
■690 ▼a0710
■71020▼aGeorgia Institute of Technology.
■7730 ▼tDissertations Abstracts International▼g87-05A.
■790 ▼a0078
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17360438▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
