The experimental investigation and validation of predictions of the Standard Model of Particle Physics is of fundamental importance for the theory of matter and affects deep questions on our understanding of nature. A key challenge in experimental studies is the evaluation of large amounts of data and the discrimination of the signal of interest from a dominating background. This thesis develops a machine learning tool for event classification in diboson channels in the ATLAS experiment at the Large Hadron Collider at CERN. A Convolutional Neural Network is trained to discriminate electroweak $WZjj eg EW$ signal from a $WZjj eg QCD$ background – demonstrating the capabilities of the new method on a use case that is of major importance for testing the Standard Model in general and the electroweak theory in particular. The developed method uses supervised deep learning to analyze event data represented in a matrix format termed event matrix while operating on highly ecient data flow graphs. By implementing a CNN for event classification, a novel method for the analysis of event data is introduce and a potential setup as well as an overview of challenges for utilizing deep learning in experimental studies on collision data is given.