Kaur, Damandeep2025-08-282025-08-282025https://hdl.handle.net/1828/22677Electron charge misidentification constitutes a significant background in analyses involving same-sign electron pairs, such as searches for electroweak production of same sign W±W± bosons. An estimation of this background is essential for improving signal sensitivity in such processes. This thesis presents the derivation of charge misidentification scale factors using a Deep Neural Network (DNN) based electron identification (ID) in the ATLAS experiment, utilizing proton–proton collision data produced at √s = 13 TeV and √s = 13.6 TeV. A data-driven method based on the Z → e+e− process is employed to estimate charge flip probabilities in data and Monte Carlo simulation, across different kinematic bins of transverse momentum and pseudorapidity. The DNN-based ID algorithm offers improved discrimination power compared to traditional likelihood-based methods, particularly in complex detector regions. The derived scale factors correct for mismodelling of charge flip rates in Monte Carlo simulations and are parametrized in both one-dimensional and two-dimensional schemes. Closure tests are performed to validate the robustness of the scale factors and their applicability across various physics analyses.enAvailable to the World Wide WebThesis