Investigation of a new approach for truth tagging within the $t$$\bar{t}$H($bb$) analysis via a graph neural

In this thesis, a new approach for truth tagging within the $t$$\bar{t}$H($bb$) analysis via a graph neural network is analysed. It is analysed how the neural network performs in comparison to the established efficiency map and the direct tag for various jet regions, working points and number of tagged jets. Furthermore, the influence of hadron variables on the performance of the neural network and the ability of the neural network to predict bottom-jet multiplicities is investigated. It is shown that the neural network can deal with low statistics and can produce similar results for various numbers of jets within an event. Furthermore it is shown that the neural network has a problem dealing with light jets which have a $p$$_{\tau}$ below 60 GeV. The cause of this problem has not been found. Additionally, it is shown that it is likely that the neural network can be used for a possible $t$$\bar{t}$H($bb$) analysis region, but more investigations are required for a final conclusion.