Applied Physics and Mathematics Annotation << Back Network model of cluster analysis Khachumov M.V. The solution of geometric clustering based on algorithm of modified artificial neural network with a different set of metrics and methods of initial placement of the clusters is proposed. The formulation of the problem of clustering is given. We consider the structure of the network model with additional memory to store the coefficients of the covariance matrix of the generalized metric. We detail the general scheme of geometric clustering based on the network model, which includes the initial placement of the clusters, setting weights of the neural network, the calculation of the covariance matrix and classification. Spherical (spatial) cluster placement model is preferred because of its simplicity and effectiveness. Traveling salesman problem proposed as a testing ground for the analysis of the network model. An algorithm to adjust the network model for solving the traveling salesman problem is given. Experimental research is made and the basic steps of the solution are graphically illustrated. The quality of network solutions is compared with the best known solutions of this problem for various cities. Evaluation of the results and recommendations for the use of the generalized metric are given. Keywords: Network model, cluster analysis, neural network, generalized metric, the covariance matrix, the traveling salesman problem, experiment, recommendations. Contacts: Email: khmike@inbox.ru Pp. 80-86.