For the construction of bottom infrastructure facilities and safe drilling on the shelf, it is necessary to conduct geological and geophysical studies aimed at studying the upper part of the section and detecting geohazardous processes. Within the framework of this work, specialized processing of 3D standard marine seismic data was performed using MASW (Multichannel Analysis of Surface Waves) technology using neural networks in order to study in detail the upper part of the section of the Pechora Sea shelf. As a result of the work done, it was possible to demonstrate the efficiency of using neural networks for processing a large amount of seismic data, which made it possible to increase the horizontal resolution of shear wave velocity models; select the optimal architecture of neural networks, and develop a methodology for training models; clarify the position of some types of geohazardous processes (paleoincisions, coarse-grained sediments) and obtain additional information on their properties and depth position.