서브메뉴
검색
Advancing Dairy Health Management Through Integrated Sensor Technologies and Machine Learning
Advancing Dairy Health Management Through Integrated Sensor Technologies and Machine Learning
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211152921
- ISBN
- 9798384053712
- DDC
- 636
- 서명/저자
- Advancing Dairy Health Management Through Integrated Sensor Technologies and Machine Learning
- 발행사항
- [Sl] : Cornell University, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 209 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-03, Section: B.
- 주기사항
- Advisor: Giordano, Julio.
- 학위논문주기
- Thesis (Ph.D.)--Cornell University, 2024.
- 초록/해제
- 요약The research presented in this dissertation aimed to enhance dairy health management by integrating sensor-based automated health monitoring (AHM) technologies and machine learning (ML) algorithms. Data from AHM systems and other precision livestock farming (PLF) technologies were used to develop health monitoring and management programs and predictive models for early lactation disease detection in dairy cows.Chapter II describes a randomized controlled trial that compared a strategy for identifying lactating Holstein cows with health disorders (HD) based on high-intensity clinical monitoring (HIC-M) versus a program that relied primarily but not exclusively on automated monitoring (AUT-M) including rumination, physical activity, and milk yield alerts. Although the HD diagnosis risk was reduced for the AUT-M program, key herd performance outcomes including milk yield, proportion of cows that left the herd, and first service reproductive outcomes did not differ for the AUT-M and HIC-M program.Chapter III explored the development and application of a framework for developing and testing machine learning algorithms for prediction of dairy cow health status using individual cow sensor and non-sensor behavioral, physiological, and performance data. Individual cow data were also combined with herd level management and environmental data. A total of 30 algorithms including non-deep learning models included in the AutoML tool Lazy Predict and deep learning models including neural networks (NN) were evaluated. The top 8 algorithms were selected for further refinement. The ensemble classifiers XGBoost (XGB) and Adaboost consistently outperformed other algorithms including NN models. Ultimately, XGB had the highest performance scores, in addition to the ability of handling missing and non-standardized data. Hyperparameter tuning was critical for improving model performance.Chapter IV focused on further developing and validating the XGB model through a workflow including feature engineering, feature selection, model training, and optimization using grid search and resampling techniques. After tunning using standard methods and domain expertise, the final XGB model presented high performance across most metrics used to measure predictive ability except precision.In conclusion, a health monitoring strategy that relied primarily on AHM systems was effective for identifying lactating dairy cows for clinical examination and ML models trained on integrated data from AHM systems and other PLF technologies were effective for predicting cow health. This research highlights the potential of sensor-based technologies to enhance dairy herd health monitoring and management which contributes to improving the sustainability of the dairy industry.
- 일반주제명
- Animal sciences
- 일반주제명
- Epidemiology
- 일반주제명
- Computer science
- 키워드
- Computer systems
- 키워드
- Dairy industry
- 키워드
- Machine learning
- 기타저자
- Cornell University Animal Science
- 기본자료저록
- Dissertations Abstracts International. 86-03B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017164191
■00520250211152921
■006m o d
■007cr#unu||||||||
■020 ▼a9798384053712
■035 ▼a(MiAaPQ)AAI31489371
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a636
■1001 ▼aPerez, Martin Matias.▼0(orcid)0000-0002-9626-8735
■24510▼aAdvancing Dairy Health Management Through Integrated Sensor Technologies and Machine Learning
■260 ▼a[Sl]▼bCornell University▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a209 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-03, Section: B.
■500 ▼aAdvisor: Giordano, Julio.
■5021 ▼aThesis (Ph.D.)--Cornell University, 2024.
■520 ▼aThe research presented in this dissertation aimed to enhance dairy health management by integrating sensor-based automated health monitoring (AHM) technologies and machine learning (ML) algorithms. Data from AHM systems and other precision livestock farming (PLF) technologies were used to develop health monitoring and management programs and predictive models for early lactation disease detection in dairy cows.Chapter II describes a randomized controlled trial that compared a strategy for identifying lactating Holstein cows with health disorders (HD) based on high-intensity clinical monitoring (HIC-M) versus a program that relied primarily but not exclusively on automated monitoring (AUT-M) including rumination, physical activity, and milk yield alerts. Although the HD diagnosis risk was reduced for the AUT-M program, key herd performance outcomes including milk yield, proportion of cows that left the herd, and first service reproductive outcomes did not differ for the AUT-M and HIC-M program.Chapter III explored the development and application of a framework for developing and testing machine learning algorithms for prediction of dairy cow health status using individual cow sensor and non-sensor behavioral, physiological, and performance data. Individual cow data were also combined with herd level management and environmental data. A total of 30 algorithms including non-deep learning models included in the AutoML tool Lazy Predict and deep learning models including neural networks (NN) were evaluated. The top 8 algorithms were selected for further refinement. The ensemble classifiers XGBoost (XGB) and Adaboost consistently outperformed other algorithms including NN models. Ultimately, XGB had the highest performance scores, in addition to the ability of handling missing and non-standardized data. Hyperparameter tuning was critical for improving model performance.Chapter IV focused on further developing and validating the XGB model through a workflow including feature engineering, feature selection, model training, and optimization using grid search and resampling techniques. After tunning using standard methods and domain expertise, the final XGB model presented high performance across most metrics used to measure predictive ability except precision.In conclusion, a health monitoring strategy that relied primarily on AHM systems was effective for identifying lactating dairy cows for clinical examination and ML models trained on integrated data from AHM systems and other PLF technologies were effective for predicting cow health. This research highlights the potential of sensor-based technologies to enhance dairy herd health monitoring and management which contributes to improving the sustainability of the dairy industry.
■590 ▼aSchool code: 0058.
■650 4▼aAnimal sciences
■650 4▼aEpidemiology
■650 4▼aComputer science
■653 ▼aComputer systems
■653 ▼aDairy industry
■653 ▼aDigital agriculture
■653 ▼aHealth management
■653 ▼aMachine learning
■653 ▼aPrecision livestock farming
■690 ▼a0475
■690 ▼a0766
■690 ▼a0984
■690 ▼a0800
■71020▼aCornell University▼bAnimal Science.
■7730 ▼tDissertations Abstracts International▼g86-03B.
■790 ▼a0058
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17164191▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
