서브메뉴
검색
Computational and Machine Learning Algorithms Development for Epidemic Forecasting and Tobacco Research
Computational and Machine Learning Algorithms Development for Epidemic Forecasting and Tobacco Research
Detailed Information
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105218
- ISBN
- 9798291566053
- DDC
- 004
- 저자명
- Tan, Jiale.
- 서명/저자
- Computational and Machine Learning Algorithms Development for Epidemic Forecasting and Tobacco Research
- 발행사항
- [Sl] : University of Michigan, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 140 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-02, Section: B.
- 주기사항
- Advisor: Eisenberg, Marisa;Meza, Rafael.
- 학위논문주기
- Thesis (Ph.D.)--University of Michigan, 2025.
- 초록/해제
- 요약Conducting research on epidemiological questions can be challenging due to the complexity of the data and underlying mechanisms involved. Mathematical modeling is crucial to public health research, in part because they allow us to extrapolate from data to provide useful insights into epidemiological patterns and trends. In this dissertation, we used and developed modeling approaches in two contexts: tobacco use and infectious disease epidemiology. We utilized a weighted Markov multistate transition model to estimate transitions in cigarette and ENDS use among youth in the US, addressing a critical public health question: Can electronic nicotine delivery systems (ENDS) be considered a harm reduction tool or are they a catalyst for cigarette use? Furthermore, we investigated this question from a biomarker perspective by employing quantile regression to examine changes in biomarkers of tobacco exposure (BOE) when cigarette smokers switch to ENDS. Finally, we applied penalized regression and covariate selection methods to integrate parameter identifiability and model selection directly into a parameter estimation process for ordinary differential equation (ODE) models, using infectious disease models as a case study. Firstly, we applied a weighted Markov multistate transition model to analyze tobacco use transitions among 28,262 observations of youth aged 12-17 from waves 2-4 (approximately 2014-2017) of the Population Assessment of Tobacco and Health (PATH) study. Our focus was on transitions between several usage states: never use, non-current experimental use, non-current regular use, current experimental use, and current regular use. We discovered that non-current experimental ENDS use is significantly positively associated with cigarette initiation. Furthermore, we observed weak positive associations between ENDS use and both cigarette progression and relapse. Conversely, while most instances of cigarette cessation do not seem to be linked to ENDS use, there is a non-significant positive association between current regular ENDS use and cigarette cessation among experimental users. Secondly, we examined whether ENDS use reduces harm by analyzing changes in biomarker levels using quantile regression. While cigarette use was linked to increases in all TSNAs and PAHs, ENDS use was associated only with certain TSNAs (e.g., NATT), suggesting potential harm reduction for biomarkers like NNAL, NABT, and NNNT. No significant link between ENDS use and PAHs was found, likely due to confounding factors. Finally, we have developed an algorithm-based framework, LASSO-ODE, designed to select identifiable mechanistic models, with a focus on infectious disease models. Our results demonstrate that the LASSO-ODE framework is highly effective at selecting a parsimonious and identifiable model from a set of larger, more complex models that may be unidentifiable, even when working with realistically sparse data containing only a single measured compartment and multiple latent (unobserved) variables. This capability is beneficial for public health practitioners and policymakers, as it supports informed decision-making and facilitates the creation of what-if scenarios, intervention testing and counterfactual analyses. Furthermore, the new cross-validation techniques for time series data introduced in our work offer promising new tools for data scientists in a variety of applications.
- 일반주제명
- Computer science
- 일반주제명
- Epidemiology
- 일반주제명
- Public health
- 키워드
- Algorithms
- 키워드
- Tobacco use
- 기타저자
- University of Michigan Epidemiological Science
- 기본자료저록
- Dissertations Abstracts International. 87-02B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017359815
■00520260202105218
■006m o d
■007cr#unu||||||||
■020 ▼a9798291566053
■035 ▼a(MiAaPQ)AAI32271785
■035 ▼a(MiAaPQ)umichrackham006338
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a004
■1001 ▼aTan, Jiale.
■24510▼aComputational and Machine Learning Algorithms Development for Epidemic Forecasting and Tobacco Research
■260 ▼a[Sl]▼bUniversity of Michigan▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a140 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-02, Section: B.
■500 ▼aAdvisor: Eisenberg, Marisa;Meza, Rafael.
■5021 ▼aThesis (Ph.D.)--University of Michigan, 2025.
■520 ▼aConducting research on epidemiological questions can be challenging due to the complexity of the data and underlying mechanisms involved. Mathematical modeling is crucial to public health research, in part because they allow us to extrapolate from data to provide useful insights into epidemiological patterns and trends. In this dissertation, we used and developed modeling approaches in two contexts: tobacco use and infectious disease epidemiology. We utilized a weighted Markov multistate transition model to estimate transitions in cigarette and ENDS use among youth in the US, addressing a critical public health question: Can electronic nicotine delivery systems (ENDS) be considered a harm reduction tool or are they a catalyst for cigarette use? Furthermore, we investigated this question from a biomarker perspective by employing quantile regression to examine changes in biomarkers of tobacco exposure (BOE) when cigarette smokers switch to ENDS. Finally, we applied penalized regression and covariate selection methods to integrate parameter identifiability and model selection directly into a parameter estimation process for ordinary differential equation (ODE) models, using infectious disease models as a case study. Firstly, we applied a weighted Markov multistate transition model to analyze tobacco use transitions among 28,262 observations of youth aged 12-17 from waves 2-4 (approximately 2014-2017) of the Population Assessment of Tobacco and Health (PATH) study. Our focus was on transitions between several usage states: never use, non-current experimental use, non-current regular use, current experimental use, and current regular use. We discovered that non-current experimental ENDS use is significantly positively associated with cigarette initiation. Furthermore, we observed weak positive associations between ENDS use and both cigarette progression and relapse. Conversely, while most instances of cigarette cessation do not seem to be linked to ENDS use, there is a non-significant positive association between current regular ENDS use and cigarette cessation among experimental users. Secondly, we examined whether ENDS use reduces harm by analyzing changes in biomarker levels using quantile regression. While cigarette use was linked to increases in all TSNAs and PAHs, ENDS use was associated only with certain TSNAs (e.g., NATT), suggesting potential harm reduction for biomarkers like NNAL, NABT, and NNNT. No significant link between ENDS use and PAHs was found, likely due to confounding factors. Finally, we have developed an algorithm-based framework, LASSO-ODE, designed to select identifiable mechanistic models, with a focus on infectious disease models. Our results demonstrate that the LASSO-ODE framework is highly effective at selecting a parsimonious and identifiable model from a set of larger, more complex models that may be unidentifiable, even when working with realistically sparse data containing only a single measured compartment and multiple latent (unobserved) variables. This capability is beneficial for public health practitioners and policymakers, as it supports informed decision-making and facilitates the creation of what-if scenarios, intervention testing and counterfactual analyses. Furthermore, the new cross-validation techniques for time series data introduced in our work offer promising new tools for data scientists in a variety of applications.
■590 ▼aSchool code: 0127.
■650 4▼aComputer science
■650 4▼aEpidemiology
■650 4▼aPublic health
■653 ▼aMathematical modeling
■653 ▼aAlgorithms
■653 ▼aElectronic nicotine delivery systems
■653 ▼aOrdinary differential equation
■653 ▼aTobacco use
■690 ▼a0984
■690 ▼a0766
■690 ▼a0800
■690 ▼a0573
■71020▼aUniversity of Michigan▼bEpidemiological Science.
■7730 ▼tDissertations Abstracts International▼g87-02B.
■790 ▼a0127
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359815▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
Preview
Export
ChatGPT Discussion
AI Recommended Related Books
Подробнее информация.
- Бронирование
- не существует
- моя папка
- Первый запрос зрения
- Non-Book Loan Application
- Nighttime Book Loan Application
Available after logging in.


