Geomechanical model construction is an essential part of field development processes planning. Building a correct pore pressure model is one of the key tasks within the process of geomechanical model construction. The traditional approach to pore pressure modeling in oil and gas industry is based on the empirical analytical models usage. This approach has a number of disadvantages, which often lead to the constructed pore pressure model to be incorrect. The authors highlight two most significant disadvantages of the traditional approach: 1) a priori discrepancy between the empirical model and fundamental physical laws; 2) the impossibility of selecting such a combination of parameters of the standard analytical model, for which the resulting pressure corresponds to the entire set of actual field data (pore pressure measurements). This paper proposes a methodology for assessing the pore pressure distribution across the field, based on the usage of neural network technology. This approach potentially eliminates both of the above disadvantages from the pore pressure model building.