Utilizing data mining techniques and ensemble learning to predict development of surgical site infections in gynecologic cancer patients

<p>Surgical site infections are costly to both patients and hospitals, increase patient mortality, and are the most common form of a hospital acquired infection. Gynecological cancer surgery patients are already at higher risk of developing an infection due to the suppression of their immune system. This research leverages popular data mining techniques to create a prediction model to identify high risk patients. Implemented techniques include logistic regression, naive Bayes, recursive partitioning and regression trees, random forest, feed forward neural network, k-nearest neighbor, and support vector machines with linear kernel. Weighted stacked generalization was implemented to improve upon the individual base level model’s performance. The chosen meta level classifiers were support vector machines with linear kernel, logistic regression, and k-nearest neighbor. The result is a model that identifies high-risk patients immediately following a surgical procedure with an AUC of 0.6864, accuracy of 0.6744, sensitivity of 0.7, and specificity of 0.6728.</p>

Name thesis:degree_name

Master of Science (MS)

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Systems Science and Industrial Engineering

Year

2018

Author dc:creator

• McDonough, John R

Contributors dc:contributor

• Dr. Mohammad T. Khasawneh

• Communication and the arts
• Applied sciences
• Health and environmental sciences
• Ensemble learning
• Gynecological cancer
• Surgical site infection
• Engineering
• Operations Research, Systems Engineering and Industrial Engineering

Repository record dc:identifier

https://orb.binghamton.edu/dissertation_and_theses/33

OAI identifier oai:identifier

oai:orb.binghamton.edu:dissertation_and_theses-1054