METHOD FOR AUTOMATIC CLUSTERING OF OBJECTS Russian patent published in 2016 - IPC G06F17/30 G06N7/06 

FIELD: computer engineering.

SUBSTANCE: invention can be used in analysis and simulation of hardly formalised processes characterised by a large number of considered factors, which requires use of special methods and instruments for multidimensional analysis of different-quality information. Method of automatic clustering of objects includes formation from the initial set of classified objects of samples in the form of initial clusters; at that, the initial set is formed by identifying every object on the basis of its parameter setting the coordinate of the object in the initial set, and is considered as a training sample formed on the basis of exponential law of distribution, while data on clusters obtained at the stage of training, is registered on corresponding elements of a memory used later during successive accumulation in them of measurement information; at the stage of training, there is also determined the model of cluster K_i with the number of elements N_i, meeting the requirement for minimum risk R_Mi(α) when forming a cluster model.

EFFECT: technical result consists in improvement of serial clustering stability.

1 cl, 2 dwg