Study of a Multivariate Technique for the Search of Single Top-Quark Production with p s = 8 TeV in the CMS Experiment at CERN

Abstract

This work is presenting a study for the search of the single top-quark production in the CMS Experiment at CERN focusing on the s-channel process and muon decay mode as the final state topology, using a multivariate technique based on the Boosted Decision Trees (BDT) algorithm. The study is based on the collision data collected at 8 TeV in the CMS detector with a luminosity of 19.3 fb^(-1). The multivariate technique is utilized with an optimization procedure for understanding what are the appropriate variables to use for separation of the signal and background events. The BDT output is obtained by optimizing the choice of the input variables by iterating in a feedback loop globally sensitive to the correlation coefficients of the variables. Then, the optimized BDT discriminant is compared with the analysis which was performed without any optimization on the choice of inputs. It has been investigated that the BDT output does not reveal any significant change in the separating power as the most globally correlated variables are removed, iteratively. Therefore, reducing the variable list in this way can be advantageous since it advances our understanding for the physical meaning of the output classifier. This study is a first consideration for the optimization of the BDT analyses in the single top-quark production and in the next step, this results will be used to fit the data accounting the systematic uncertainties and extract the cross-section for the BDT discriminant obtained so far.