As an underground multi-purpose neutrino observatory with 20 kton liquid scintillator (LS), JUNO has excellent potential to first detect Diffuse Supernova Neutrino Background (DSNB). The dominant background for the DSNB is the Neutral-Current (NC) interaction of atmospheric neutrinos with C-12 nulei in LS. The final state particles of NC interactions usually contain α,p,n, in contrast to positrons from DSNB. The LS fluorescent time profile is different for different particle types. We have developed machine learning based Pulse Shape Discrimination (PSD) methods to make use of the difference in hit time distributions. In this poster, the two PSD methods developed for DSNB search will be presented: including Boosted Decision Trees (BDT) and Neural Networks (NN).