Study of the properties of expert system for seismic event classification using convolutional neural networks

The presented article brings the reader to the final part of the development of the proposed neural network expert system for automatic classification of seismic events by their seismogram. The fundamental principles of the system were described in previous publications of the series. Here are the results of its testing for a set of characteristics related to the recognized events, conditions of their occurrence, features of registration, principles of processing records and methods of their analysis. The weaknesses and strengths of the methods used are shown. Ways to improve the quality, accuracy and speed of the expert system are outlined. Possibilities of complicating the set of characteristics are proposed, the classification of records by which can be made, for example, by tectonic conditions in the earthquake source.