Precision medicine is to customize a medical model for new tools and therapies to select best treatments, being tailored to the individual patients. Subgroup selection plays an important role in precision medicine to assess the treatment effects in subgroups; it provides useful information to optimize the treatment assignment. In this study, we propose using tree-based recursive partitioning to identify patient subgroups with the enhanced treatment effect in clinical trials. Two subgroup identification strategies are presented. One is based on the Differential Effect Search (SIDES) algorithm where the subgroups are identified by maximizing the treatment effect between treatment group and control group. SIDES generates multiple candidate subgroups; it is desirable to have one single subgroup to be used for treatment assignment. We evaluate several methods to identify “optimal” subgroups from the list of subgroups identified. The second strategy is an ensemble tree-based method. For a given terminal node in a tree, the patients in that terminal node are assigned to have a score equaling to the proportion of the responders over the node size. The patient’s composite score is calculated as sum of all ensemble trees. A change-point algorithm is then applied to separate responder and non-responder subgroups. We conduct simulation experiments to evaluate these methods and compare with subgroup discovery CN2-SD and SD-Map algorithms in term of sensitivity, specificity, and accuracy.