The selection of $\eta$ mesons with a high efficiency and a high purity can be important in the formation of statistically significant invariant mass spectra in the reconstruction of short-lived particles such as $\eta' \rightarrow \pi^{+} \pi^{-} \eta$. In this study, a cut-based standard method and a ranking method to reduce combinatorial background in the reconstruction of $\eta \rightarrow \gamma \gamma$ decays in high multiplicity hadronic events are presented. By using recorded ALEPH data and fully simulated events, the performances of the methods are compared. Results show that the ranking method yields significant improvements in the purity of the selected $\eta$ meson relative to the standard method.