Artificial Immune Classifier (aiCLS): An Immune Inspired Supervised Machine Learning Method

Artificial immune systems have been proven to be efficient in pattern recognition, data clustering and dataclassification. the proposed method is a novel artificial immune classifier called aicls based on ainet. artificialimmune network (ainet) is an efficient data analysis and clustering algorithm capable of clustering simple datasetsthrough complex ones. hidden capabilities of ainet for supervised learning were significantly considered by aicls.the proposed method takes a local optimization approach to classification problem. it generates local optimum cellsto recognize any given training antigen. concatenation of these cells results in a global optimum classifier. the noveltyof aicls has been discussed from both computational and immunological aspects. from the computational aspect,aicls is a fast one-shot learner algorithm with regard to the proposed “iterative clonal selection”. from theimmunological aspect, aicls introduces a novel clonal suppression method called “dissimilarity proportional clonalsuppression (dpcs)”, which increases data reduction and convergence to local optimum for any given antigen. dpcsalters convergence through a greedy suppression, which takes antibody-antigen affinity into account. theexperimental results show that aicls outperforms artificial immune recognition system (airs) on uci benchmarkdatasets in both classification accuracy and data reduction.