Complex classification of global geomagnetic disturbances

Based on the Kohonen algorithm, a self-training neural network is constructed which allows one to classify geomagnetic disturbances using the data on parameters of the solar wind and interplanetary magnetic field. Such an approach permits one to consider the suggested classification simultaneously as space and physical, since the space origin of disturbances of different kinds is considered within the framework of the classification. As a result of numerical experiments, we have succeeded in isolating basic classes of complexes of disturbed parameters accounting for various events of the space weather, each of which is responsible for corresponding global magnetospheric conditions.