A BDT optimization study and assessment of deep learning in selecting VBF events in the $H\to ZZ^{*}\to4l$ channel

Since the discovery of the Higgs boson in 2012, the most recently confirmed and final piece of the Standard Model has been rigorously studied in various production modes and decay channels. However, Run 1 of the LHC did not yield sufficient statistics to resolve production modes in the $H\to ZZ^{*}\to4l$ channel. Current and future runs of the LHC will offer sufficient data to do so, thus machine learning techniques used to separate the two most frequent Higgs boson production modes - gluon fusion and vector boson fusion - were studied and the findings are presented in this report. An optimization of boosted decision trees trained on $\sqrt{s}=\text{8 TeV}$ ATLAS Monte Carlo data is presented. The feasibility of improving classification efficiency by using deep neural networks is also studied and detailed below.