The High Altitude Water Cherenkov (HAWC) Observatory is a ground based air-shower array of 300 water Cherenkov detectors (WCDs) located on the hillslide of volcano Sierra Negra, Mexico. Each WCD has 4 photomultiplier tubes (PMTs) that detect secondary particles of air-showers produced by gamma-rays and cosmic rays (CRs). Those CRs are the main problem in the gamma-ray sources analysis, therefore, we need to separate between both particles. Currently, the HAWC data is divided in 10 bins that depend on the number of PMTs activated in each event. For the suppression of CRs background, HAWC uses two variables, Compactness and PINCness, that are used to apply a simple cutoff in each bin. In this work a neural network (NN) was trained that uses these two variables as input parameters in order to obtain one output parameter and use it as a cutoff. We used simulated proton and gamma events to train the NN and we found an optimal cutoff, that we applied to the Crab Nebula. This work predicts a better gamma/hadron separation in some bins when we use Monte Carlo (MC) data.