Modification of the K-means Method with an Unknown Number of Classes

The classification procedure based on the K-means method with an unknown number of classes first proposed by professor Sergey Aivazian is designed and investigated in the article. The adaptive Mahalanobis metrics and dynamically computed measures of observation anomaly and class homogeneity are used in the paper. The classification of regions of The Russian Federation according to quality of life integral characteristics is obtained. Finally a brief interpretation of the results is given