Measurements of the Higgs boson using the H ! Higgs boson decay mode and two di↵erent methods for identifying Higgs bosons produced via vector boson fusion are presented. These analyses use proton-proton collision data collected by the CMS collaboration during the 2016 running period and constitute 35.9fb 1 of integrated luminosity at ps = 13 TeV. One vector boson fusion identification method is based on boosted decision trees, and the other is based on jets formulated as images and a dense convolutional neural network. The categorisations produced by both methods are subjected to the overall H ! statistical analysis and their results compared. The neural network itself is also subjected to analysis to determine what features it has learned to extract from the jet images.
The main objectives of this new approach are to reduce contamination from gluon fusion in the vector boson fusion categories and to improve their statistical signif- icance. This is indeed observed in the expected yields measured from simulation where vector boson fusion category signal purity and significance are both increased. The measurement of the vector boson fusion signal strength modifier is improved in the new approach where we observe reduced uncertainties. The value relative to the
Standard Model is measured to be 0.8+0.6 in the boosted decision tree variant and +0.5 0.5
1.5 0.5 in the neural network variant. The neural network is also observed to give a reduced uncertainty on many of the other measurements, especially those more directly impacted by vector boson fusion production.