Machine learning approach to $H$ $\rightarrow$ $\tau$$\tau$ analysis in the CMS experiment

The ATLAS and CMS experiments at the LHC discovered a 125 GeV Higgs boson in 2012. The measurements of its properties give us an insight to many important physical parameters. In this thesis the CMS H → ττ → μνμνττhντ analysis is discussed, including the Higgs physics theory, construction of the CMS apparatus and event reconstruction algorithms. Moreover, a set of machine learning methods is presented together with their utilization in event identification. The construction of these methods allows them to be easily accommodated to official Monte Carlo CMS data samples in the supervised learning mode. We showed that the result of 0.875 for average area under receiver operating curves for all considered event final states is feasible. The best performance is obtained for neural network model. The thesis contains the author analysis of significance implemented on top of both cut based and machine learning based distributions of discriminating variables in the aforementioned Higgs boson decay channel. The results indicate that the incorporation of the output of machine learning estimator can boost the performance of the analysis by 20 to 35%.