<p>It is well known now that in diseases like cancer where the disease progression varies from patient to patient, no single treatment benefits all patients. With the advancement of medical science in the past two decades, targeted therapies based on individualized genomic profiles have been developed to treat different types of cancer patients. The development of customized therapies, often referred to as personalized or stratified medicine, is poised to revolutionize drug development. With the central goal of customizing medical care and cures, the focus of personalized medicine is to administer the right drug with the right dose to the patient recognized through a large number of patient characteristics (several clinical features and biomarkers, etc.). The natural demand for appropriate statistical techniques arises to find effective tools for the construction and evaluation of evidence-based personalized intervention strategies. In the present discussion, we emphasize our study on the direction of tracking the right group of patients given a specific treatment. This leads to appropriate subgroup identification. In fact, we propose a subgroup identification technique for clustered data using a supervised learning approach. To the best of our knowledge, only a couple of articles are available for model based (MOB) partitioning with clustered data. We develop a semiparametric model based partitioning approach called Quadratic Inference Function Tree (QUIFT). The newly proposed approach is very much comparable with the existing fully parametric MOB tree. The performance of the proposed technique is evaluated using simulation studies. The technique has been implemented for the analysis of a real data set obtained from a multi-center clinical trial.</p>