|
Abstract:
|
Our purpose is to evaluate and compare the performances of different classification methods for predicting prostate cancer outcomes using gene expression data. Specifically, we develop a systematic statistical strategy for constructing a reliable and precise classifier for predicting cancer outcome. We focus on comparing selected statistical methods to predict binary cancer outcome using gene expression data from a cohort of Swedish prostate cancer patients. The methods we compare are logistic regression, lasso regression, regression trees, random forests, gradient-boosted machines, and support vector machines. We perform extensive simulation studies to compare the performance of the selected classification methods.
|
