Skip to Main Content
Frequently Asked Questions
Submit an ETD
Global Search Box
Need Help?
Keyword Search
Participating Institutions
Advanced Search
School Logo
Files
File List
Master_s_Thesis_16.pdf (1.1 MB)
ETD Abstract Container
Abstract Header
A Modified Q-Learning Approach for Predicting Mortality in Patients Diagnosed with Sepsis
Author Info
Dunn, Noah M
ORCID® Identifier
http://orcid.org/0000-0002-9419-7953
Permalink:
http://rave.ohiolink.edu/etdc/view?acc_num=miami1618439089761747
Abstract Details
Year and Degree
2021, Master of Computer Science, Miami University, Computer Science and Software Engineering.
Abstract
Among the medical crises of modern times, medically-diagnosed Sepsis persists as an ongoing condition yielding high mortality across all spectra of patients. Patients with Sepsis suffer variable symptoms, making it hard to evaluate severity, and the outcome of patients who develop Sepsis can range from full-recovery to death. The FOOTON model proposed in this study (named for its use of the qSOFA and SOFA scoring metrics) seeks to aid physicians in evaluating patient condition severity. The FOOTON model makes use of a stratified, cross-validated version of data provided by the MIMIC-III dataset to construct four variants of a binary classification model. Upon the completion of the model construction, the models can take patient data as input and they will output their prediction on the binary outcome of the patient (life or death). Of the four models produced, two models were shown to have the most promise. The imbalanced, unweighted variant of the model which performs at an average accuracy of 78.413% overall, and a balanced weighted variant, which performs at an average accuracy of 61.638% for dead patients. Providing this capability as an assistive tool for physicians can allow for the prioritization of limited resources to individuals at a greater risk of dying, with the potential to decrease overall patient mortality.
Committee
Dhananjai Rao (Advisor)
Philippe Giabbanelli (Committee Member)
Vaskar Raychoudhury (Committee Member)
Pages
69 p.
Subject Headings
Computer Science
Keywords
Sepsis, Machine Learning, Q-Learning, MDP, ICU, MIMIC-III, Mortality, In-Hospital, Noah Dunn, Model, Reinforcement Learning
Recommended Citations
Refworks
EndNote
RIS
Mendeley
Citations
Dunn, N. M. (2021).
A Modified Q-Learning Approach for Predicting Mortality in Patients Diagnosed with Sepsis
[Master's thesis, Miami University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=miami1618439089761747
APA Style (7th edition)
Dunn, Noah.
A Modified Q-Learning Approach for Predicting Mortality in Patients Diagnosed with Sepsis.
2021. Miami University, Master's thesis.
OhioLINK Electronic Theses and Dissertations Center
, http://rave.ohiolink.edu/etdc/view?acc_num=miami1618439089761747.
MLA Style (8th edition)
Dunn, Noah. "A Modified Q-Learning Approach for Predicting Mortality in Patients Diagnosed with Sepsis." Master's thesis, Miami University, 2021. http://rave.ohiolink.edu/etdc/view?acc_num=miami1618439089761747
Chicago Manual of Style (17th edition)
Abstract Footer
Document number:
miami1618439089761747
Download Count:
240
Copyright Info
© 2021, some rights reserved.
A Modified Q-Learning Approach for Predicting Mortality in Patients Diagnosed with Sepsis by Noah M Dunn is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License. Based on a work at etd.ohiolink.edu.
This open access ETD is published by Miami University and OhioLINK.