Performance Evaluation of Multivariate Analysis Methods on the $Z \gamma$ Final State

The performance of various machine learning algorithms are evaluated for their separation power of the $Z\gamma$ Electroweak process (with $Z\rightarrow\ell\ell$ and $\ell=e,\mu$) against the various backgrounds that populate the selection.\\ The Boosted Decision Tree method is found to deliver the best performance and is compared to that of neural net analysis and previously used methods using $36.1\, \text{fb}^{-1}$ of data obtained at $\sqrt{s}=13\, \text{TeV}$ from the ATLAS detector in 2015 and 2016.