In the Standard Model (SM) the Higgs mechanism gives vector bosons (VB) their mass via the electroweak symmetry breaking (EWSB). Massive VB have the possibility of being longitudinally polarized. Because of that, the existence of massive longitudinal polarized VB is closely linked to the Higgs mechanism. In order to study the longitudinal polarization, the scattering of two massive gauge bosons is looked at. Since the longitudinal polarization component is rare compared to its background, an important part of the analysis is the identification of this polarization component. For this task Neural Networks (NN) are used. In order to improve the quality of the used NNs, a framework called OPTIMA is to perform a hyperparameter optimization. In this thesis the successful implementation of Lightning in OPTIMA and the exploration of different possibilities to use multiclass classifier and regression models to find the longitudinal polarization component in the scattering of two same-charged W± bosons is described. The tested possibilities show that for the discrimination of the longitudinal polarization component, multiclass classifier outperform regression models. However, from the observed behaviour it seems promising to combine multiclass classifier and regression model in order to achieve better performances. Furthermore, it is shown that performing the SR event selection after training the NN instead of before, results in a significantly better performance on the SR. Another result of this thesis is that using the same events multiple times but with different target labels also improves the performance of the NN significantly. However, most of these effects are dependent on the size of the available dataset and therefore need further testing with differently sized datasets.