The next generation water-Cherenkov detector, Hyper-Kamiokande (Hyper-K), is currently under construction in Japan and it is expected to be ready for data taking in 2027. For its huge fiducial volume and high statistics, Hyper-K will contribute to many investigations such as CP-violation, determination of neutrino mass ordering, and potential observations of neutrinos from astrophysical sources. To optimize the sensitivity of the detector, Hyper-K will have a hybrid configuration of photodetectors: thousands of 20-inch photomultipler tubes (PMTs) will be combined with modules containing smaller PMTs arranged inside a pressure vessel, called multi-PMT modules. Many efforts are on-going to reduce the expected dark counts from both photodetectors. In this poster, we report the details and performances of multivariate analysis techniques such as Boosted Decision Tree that are currently being applied to simulated events to reduce the overall dark rates of the detector, which is significantly important for Hyper-K's sensitivity to low energy neutrinos.