Ruoqing Zhu, University of  Illinois Urbana-Champaign

"Tree-based methods in personalized medicine"

Tree-based methods provide a flexible framework for classification, regression and survival models. Ensemble tree models provide accurate predictions, while single tree models are more interpretable. In this talk, I will present two of our recent developments of tree-based methods for personalized medicine, both under the outcome weighted learning framework. The first approach is a single tree model with a greedy learning of splitting rules to learn the optimal treatment rule. This approach is more likely to produce simpler tree structures with more accurate decision rules than existing methods. The second approach deals with censored survival outcome and utilizes an imputation of the censored observations based on survival tree ensembles. The imputation avoids the estimation of the inverse probability of censoring, which is oftentimes unstable. We demonstrate the performance of the proposed methods through both simulation studies and real data analysis.

Stay connected TwitterFacebookLinkedInYouTubeInstagram