Classification of Objects Based on a Tree-Shaped Artificial Immune Network Model

Abstract

This paper describes the solution of the object classification problem in a multidimensional attribute space based on a modified artificial immune network model using the principles of a minimum spanning tree (MST). During the classification with unsupervised learning (clustering) at different execution stages of the immune network, objects are used as antigens and antibodies to form a training sample (classifier). In the case of classification with supervised learning, objects from a training set are used as a set of antigens, and objects to classification are used as a set of antibodies. The class definition for each object is based on the avidity value, which describes the strength of cooperative affinity interactions of antibodies with antigen. Using the proposed model allows to speed up the classification process in comparison with models based on the MST and C-means methods, as well as automating the process of determining the number of classes in the absence of a training sample.KeywordsClassificationClusteringImmune learningAvidityValenceTree-based immune network