Measurement of the $\text{t}\overline{\text{t}}\text{H}$($\text{b} \overline{\text{b}}$) process in the dilepton final state with machine learning techniques at the CMS experiment

This thesis presents the first measurement of the associated production of a Higgs boson and a top-quark-antiquark pair (ttH) in final states with two leptons and where the Higgs boson decays into a bottom-quark-antiquark pair (H ! bb). The measurement utilizes proton-proton collision data collected at the CERN LHC with the CMS experiment at a center-of-mass energy ps=13 TeV during 2016-2018 of Run 2, which corresponds to an integrated luminosity of 138 fb1 . In the Standard Model (SM), the Higgs boson couples to elementary particles with a Yukawa-type interaction and a coupling strength proportional to the particle mass. Since the top quark is the heaviest elementary particle known to date, the ttH process is the best direct probe of the top-Higgs Yukawa coupling and, therefore, a vital element to verify the SM nature of the Higgs boson. Furthermore in the SM, the H ! bb decay presents the largest branching fraction and is thus experimentally attractive as a final state. The dominant background contributions arise from tt events containing additional jets that do not originate from the tt system (tt + jets). In particular, tt + bb processes constitute an almost irreducible background with respect to the ttH, H ! bb signal. To improve the sensitivity of the measurement, dedicated multivariate analysis techniques are used to extract the signal. This study includes detailed assessments of the performance and optimization of discriminants trained using, for the first time, Artificial Neural Networks, as well as improved modeling of the dominant background contributions. The results are interpreted in terms of the inclusive ttH signal-strength modifier (µttH) and the first ttH, H ! bb STXS measurement, as a powerful tool to test the SM predictions and search for new physics. The statistical model used for measurements effectively describes well the observed data. The observed best fit value is µttH = 0.34+0.17 0.17(stat.) +0.20 0.21(syst.) for the combination of all decay channels. Additionally, this thesis presents an overview of the luminosity measurement for the CMS experiment calibrated using the Van der Meer method. The focus is on an improved method used to estimate the bias due to the assumption of factorizable proton densities of the beams using transverse densities determined from beam-imaging scan data, resulting in a significant reduction of the systematic uncertainty for the measurement of the integrated luminosity