While Western countries typically run census surveys frequently, poorer countries such as Haiti do not have the money to do so; thus research into how Haitians live is severely lacking. Furthermore, studies that do exist tend to be not only old and outdated, but also lacking in depth. Using new census data recently collected from Haiti, I attempt to predict if certain behaviors and living situations can be used as indicators for determining if someone has cholera. Challenges for exploring this data center on getting the surveys into a format suitable for analysis and the severe class imbalance between the number of cholera positive people and cholera negative people. Numerous solutions to this problem are attempted including using different sampling techniques, using ensembles with models like CART and SVM, and Bayesian model averaging. Better survey designs and questions to add to future surveys are also discussed.
Applied Statistics (MS)
Department, Program, or Center
School of Mathematical Sciences (COS)
Young, Jessica, "Predicting Cholera Positive Cases in Haiti" (2017). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus