Кластерізація потоків даних з використанням самоорганізовних карт Т. Кохонена

Abstract

This thesis is devoted to clustering of multidimensional data using T. Kohonen`s self-organizing maps. Today amount of data increases each day. The problem is that it becomes almost impossible too process and cluster this quantity of multidimensional data manually. That`s why we use clustering. Main idea of self-organizing maps is to produce a low-dimensional (typically twodimensional), discretized representation of input values. This artificial neural network is trained unsupervised learning and apply competitive learning as opposed to error-correction learning. Competitive learning is achieved by using a neighborhood function to preserve the topological properties of the input space.