Constructing a Predictive Model for the Winner of Survivor

Throughout this paper, we discuss various predictive models that could be used to predict the winner of CBS's reality television show, Survivor, which is hosted by Jeff Probst. We first give an in-depth explanation to how the data were collected and sorted, and what the variables in the data mean. We then apply a series of predictive models to the data and analyze the results in order to determine whether the winner of Survivor can be predicted based on information the audience knows prior to the merge. If a model under consideration does not work, we explain why it fails. For the predictive model that we eventually propose for the show, we first apply Principal Component Analysis in order to achieve dimension reduction on the number of continuous variables of the collected data and then quantize them to construct a Naive Bayes' Classifier model along with other categorical variables.