B1627
Title: Mining personalized treatment effects by gradient boosting tree
Authors: Shonosuke Sugasawa - Keio University (Japan) [presenting]
Hisashi Noma - The Institute of Statistical Mathematics (Japan)
Abstract: It has been recognized that treatment effects of the same treatment are often heterogeneous among patients, and mining the personalized treatment effects is essential for achieving personalized and precision medicine. Here we propose an effective way to estimate the personalized treatment effects by using the gradient boosting tree method known as a powerful nonparametric regression method in machine learning field. We estimate relationships between outcome and individual covariates in treatment and control (alternative) groups, respectively, in randomized clinical trials, and combine them to obtain the personalized treatment effect estimates. We apply the proposed method together with some existing methods to simulated data set and real trial data.
