The Super-Kamiokande experiment (Super-K) is a water Cherenkov detector noted for its discovery of the oscillation of atmospheric neutrinos. The dominant effect of the oscillation of muon neutrinos in the atmosphere is the appearance of tau neutrinos. Direct detection of $\nu_\tau$ in the atmospheric neutrino flux would provide a clear confirmation of neutrino oscillations. The sub-dominant $\nu_\mu$ oscillation mode, of $\nu_\mu$ changing to $\nu_e$, is studied at Super-K to determine mass ordering. Currently, $\nu_\tau$ interactions form the biggest background to the mass ordering signal in the Super-K analysis. Machine learning techniques of neural networks are used at Super-K to segregate $\nu_\tau$ charged-current interactions from the interactions of the atmospheric muon and electron neutrinos. 10% more events can be added to the analysis by expanding the fiducial volume of the detector. Studies on the Monte-Carlo simulations for a Super-K run period, between 2008 to 2018, suggest that we can expect improvements in the search for tau neutrinos and the suppression of mass-ordering backgrounds.