In this article, an analysis of the Higgs boson production via vector boson fusion in the SM H→WW→ 2l2ν (l = e, μ) is performed from an optimization technique in the event selection, called DNN analysis. This analysis compares the standard selection process that CERN performs to study the production of a particle from a cut-based analysis, where the study of statistical significance shows that DNN analysis can better separate signal and background events. To perform the DNN analysis, we optimized the neural network configuration to discriminate signal and background events effectively. Moreover, studies of activation functions such as RELU and Sigmoid, stochastic optimization methods such as ADAM, and regularization methods such as Dropout. All this leads to constructing an optimal neural network topology capable of learning events and signal and background discrimination. Finally, we found an important improvement of approximately 47% and 27% for Z_VBF and Z_Higgs, respectively.
