Multivariate analysis of vector-boson-fusion di-Higgs production in the bb̅gg final state using XGBoost

Abstract

A multivariate analysis using XGBoost is performed for the vector-boson-fusion di-Higgs production in the final state consisting of a bottom−antibottom quark pair and a gluon pair. Events are first selected using an efficiency-based tagging strategy to define the analysis dataset. The classifier combines jet kinematic and substructure observables into an event-level discriminant, which is used to define an optimal selection that maximizes the expected signal significance. A significance of Z ≈ 0.43 is achieved in this challenging hadronic final state. The performance is evaluated using a nested cross-validation procedure.