An Adaptive Artificial Immune Network Classifier with Independent Suppression Threshold

Abstract

Artificial immune algorithm (aiNet) is one of the algorithms of the artificial immune system that is introduced as clustering and filtering of redundant data. This algorithm is also used as a classifier. One of the most effective parameters in this network is the suppression threshold which is responsible for controlling the value of minimum distance between two antibodies in the training phase, and the recognition threshold between antibody and antigen in the testing phase. The efficiency of the results of aiNet algorithm depends on the suppression threshold parameter and due to the fact that the suppression threshold parameter depends on the input data, in this paper we introduce the vectorial suppression threshold parameter (vts) instead of suppression threshold in order to automatic tuning of this parameter. We present an adaptive system, based on the feedback system which is capable of adjusting separate value of the suppression threshold for each class. The proposed method is tested on UCI dataset and Corel image dataset. The results show that the proposed model is acceptably effective and advantageous in comparison with the base method and other classifier.